Assignment 2: Exploring Tree-Based
Regression Methods f

Question

Assignment 2: Exploring Tree-Based
Regression Methods for 2D Linear Data

TutorBin · Accepted Answer

{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Assignment 2: Exploring Tree-Based Regression Methods for 2D Linear Data ", "## DTSC 680: Applied Machine Learning ", " ", "## Name: " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Directions and Overview ", " ", "The main purpose of this assignment is for you to gain experience using tree-based methods to solve simple regression problems. In this assignment, you will fit a `Gradient-Boosted Regression Tree`, a `Random Forest`, and a `Decision Tree` to a noisy 2D linear data set. Since these models can be trained very quickly on the supplied data, I want you to first manually adjust hyperparameter values and observe their influence on the model's predictions. That is, you should manually sweep the hyperparameter space and try to hone in on the optimal hyperparameter values, again, _manually_. (Yep, that means guess-and-check: pick some values, train the model, observe the prediction curve, repeat.) ", " ", "But wait, there's more! Merely attempting to identify the optimal hyperparameter values is not enough. Be sure to really get a visceral understanding of how altering a hyperparameter in turn alters the model predictions (i.e. the prediction curve). This is how you will build your machine learning intuition! ", " ", "So, play around and build some models. When you are done playing with hyperparameter values, you should try to set these values to the optimal values manually (you're likely going to be _way_ off). Then, retrain the model. Next in this assignment, we will perform several grid searches, so you'll be able to compare your "optimal" hyperparameter values with those computed from the grid search. ", " ", "We will visualize model predictions for the optimal `Gradient-Boosted Regression Tree`, a `Random Forest`, and `Decision Tree` models that were determined by the grid searches. Next, you will compute the generalization error on the test set for the three models. ", " ", "You will think critically about the different algorithms, as well as their prediction results, and characterize the trends you observe in the prediction results. Can you explain in your own words how the algorithms work? How do the results predicted by these models compare and contrast? Can you see any relationship between the algorithms and the model results? ", " ", "Lastly, based on the arguments outlined in your critical analysis you must identify the model that best characterizes this data. In other words, which of these models is going to generalize better? Consider all factors." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Preliminaries ", " ", "Let's import some common packages:" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "# Common imports ", "import matplotlib.pyplot as plt ", "import matplotlib as mpl ", "from matplotlib import cm ", "import numpy as np ", "import pandas as pd ", "%matplotlib inline ", "mpl.rc('axes', labelsize=14) ", "mpl.rc('xtick', labelsize=12) ", "mpl.rc('ytick', labelsize=12) ", "import os ", " ", "# Where to save the figures ", "PROJECT_ROOT_DIR = "." ", "FOLDER = "figures" ", "IMAGES_PATH = os.path.join(PROJECT_ROOT_DIR, FOLDER) ", "os.makedirs(IMAGES_PATH, exist_ok=True) ", " ", "def save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300): ", " path = os.path.join(IMAGES_PATH, fig_id + "." + fig_extension) ", " print("Saving figure", fig_id) ", " if tight_layout: ", " plt.tight_layout() ", " plt.savefig(path, format=fig_extension, dpi=resolution) ", " ", "def plot_predictions(regressor, X, y, axes, label=None, style="r-", data_style="b.", data_label=None): ", " x1 = np.linspace(axes[0], axes[1], 500) ", " y_pred = regressor.predict(x1.reshape(-1, 1)) ", " plt.plot(X[:, 0], y, data_style, label=data_label) ", " plt.plot(x1, y_pred, style, linewidth=2, label=label) ", " if label or data_label: ", " plt.legend(loc="upper center", fontsize=16) ", " plt.axis(axes)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Import and Split Data ", " ", "Complete the following: ", " ", " ", " ", "1. Begin by importing the data from the file called `2DLinear.csv`. Name the returned DataFrame `data`. ", " ", "2. Call [train_test_split()](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html) with a `test_size` of 20%. Save the output into `X_train`, `X_test`, `y_train`, and `y_test`, respectively. Specify the `random_state` parameter to be `42` (do this throughout the entire note book). ", "3. Reshape some data, so that we don't run into trouble later." ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "### ENTER CODE HERE ### ", "from sklearn.model_selection import train_test_split ", " ", "data = pd.read_csv('./2DLinear.csv') ", "X, y = data['x'], data['y'] ", "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42) ", "### ENTER CODE HERE ### ", " ", "# Reshape X Data and Make NumPy Arrays ", "X_train = np.array(X_train).reshape(-1,1) ", "X_test = np.array(X_test).reshape(-1,1) ", "y_train = np.array(y_train) ", "y_test = np.array(y_test)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Plot Data ", " ", "Simply plot your training data here in order to create the plot below. Be sure you are using the training data when you plot, not the testing data." ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure Raw2DLinearData " ] }, { "data": { "image/png": "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 ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plt.scatter(X_train, y_train) ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('Noisy 2D Linear Data') ", "fig_id = 'Raw2DLinearData' ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Explore 2D Linear Data: GradientBoostingRegressor ", " ", "Fit a `GradientBoostingRegressor` model to this data. You must manually assign values to the following hyperparameters. You should "play around" by using different combinations of hyperparameter values to really get a feel for how they affect the model's predictions. When you are done playing, set these to the best values you can for submission. (It is totally fine if you don't elucidate the optimal values here; however, you will want to make sure your model is not excessively overfitting or underfitting the data. Do this by examining the prediction curve generated by your model. You will be graded, more exactly, on the values that you calculate later from performing several rounds of grid searches.) ", " ", " - `learning_rate = ` ", " - `max_depth = ` ", " - `n_estimators = ` ", " - `random_state = 42`" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Score on test set is 0.7074 " ] } ], "source": [ "### ENTER CODE HERE ### ", "from sklearn.ensemble import GradientBoostingRegressor ", "reg = GradientBoostingRegressor(learning_rate=0.1, max_depth=3, n_estimators=50, random_state=42) ", "reg.fit(X_train, y_train) ", "print('Score on test set is {:.4f}'.format(reg.score(X_test, y_test)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plot Model Predictions for Training Set" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure gbrt_2DLinearReg " ] }, { "data": { "image/png": "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 ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plot_predictions(reg, X_train, y_train, [0.0, 2.0, 2.0, 12.0], label='Ensemble Predictions') ", "fig_id = 'gbrt_2DLinearReg' ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('GradientBoostingRegressor: learning_rate=0.1, n_estimators=50') ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show() ", "### ENTER CODE HERE ###" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Explore 2D Linear Data: RandomForestRegressor ", " ", "Fit a `RandomForestRegressor` model to this data. You must manually assign values to the following hyperparameters. You should "play around" by using different combinations of hyperparameter values to really get a feel for how they affect the model's predictions. When you are done playing, set these to the best values you can for submission. (It is totally fine if you don't elucidate the optimal values here; however, you will want to make sure your model is not excessively overfitting or underfitting the data. Do this by examining the prediction curve generated by your model. You will be graded, more exactly, on the values that you calculate later from performing several rounds of grid searches.) ", " ", " - `min_samples_split = ` ", " - `max_depth = ` ", " - `n_estimators = ` ", " - `random_state = 42`" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Score on test set is 0.7049 " ] } ], "source": [ "### ENTER CODE HERE ### ", "from sklearn.ensemble import RandomForestRegressor ", "reg = RandomForestRegressor(min_samples_split=2, n_estimators=100, random_state=42) ", "reg.fit(X_train, y_train) ", "print('Score on test set is {:.4f}'.format(reg.score(X_test, y_test)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plot Model Predictions for Training Set" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure rf_2DLinearReg " ] }, { "data": { "image/png": "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 ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plot_predictions(reg, X_train, y_train, [0.0, 2.0, 0.0, 14.0], label='Ensemble Predictions') ", "fig_id = 'rf_2DLinearReg' ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('RandomForestRegressor') ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show() ", "### ENTER CODE HERE ###" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Explore 2D Linear Data: DecisionTreeRegressor ", " ", "Fit a `DecisionTreeRegressor` model to this data. You must manually assign values to the following hyperparameters. You should "play around" by using different combinations of hyperparameter values to really get a feel for how they affect the model's predictions. When you are done playing, set these to the best values you can for submission. (It is totally fine if you don't elucidate the optimal values here; however, you will want to make sure your model is not excessively overfitting or underfitting the data. Do this by examining the prediction curve generated by your model. You will be graded, more exactly, on the values that you calculate later from performing several rounds of grid searches.) ", " ", " - `splitter = ` ", " - `max_depth = ` ", " - `min_samples_split = ` ", " - `random_state = 42`" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Score on test set is 0.7001 " ] } ], "source": [ "### ENTER CODE HERE ### ", "from sklearn.tree import DecisionTreeRegressor ", "reg = DecisionTreeRegressor(splitter='best', min_samples_split=2, random_state=42) ", "reg.fit(X_train, y_train) ", "print('Score on test set is {:.4f}'.format(reg.score(X_test, y_test)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plot Model Predictions for Training Set" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure dt_2DLinearReg " ] }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAAFgCAYAAACmDI9oAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAAsTAAALEwEAmpwYAABlgUlEQVR4nO3dd3hb1f3H8ffxThxnO3sCWYQMQhIwEGJIKKMUWmgYDXsVyiq0tIz+2FBa6GAWApRNoey2bAIOy4wwAoGQELIX2YmdYSf2+f1xJPtalizJmpY/r+fRI/vqSjq6vrY/+up7zzXWWkRERERExMlK9QBERERERNKJArKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIhwKyiIiHMeYVY8wpEaxXaYzZJRljEhGR5FJAFpEWyRizyBizzRhTYYzZaIz5wBhzjjEmpr9r1trDrLUPR7BeO2vtglieyxjztS9oVxpjaowx2z3fXxHLYwc8z0PGmGrf4643xrxhjBkar8cXEck0Csgi0pL9xFpbBPQHbgZ+DzyQ2iFFzlo73Be02wHvAuf7v7fW3uRfzxiTE4en+7PveXoDy0nAdorTOGN5fhPrGyQREVBAFpEMYK3dZK39D3AccIoxZg9jTL4x5lZjzBJjzA/GmHuMMW389zHGHGWM+cIYs9kY870x5lDf8jJjzJm+r3czxswwxmwyxqw1xjzlub81xuzm+7qDMeYRY8waY8xiY8wf/EHNGHOqMeY931g2GGMWGmMOa+r1GGMG+B7/DGPMEuAt3/LTjTFzfI/zmjGmv+c+Q32V4fXGmLnGmGNDbKttwL+B0Z779jLGPOsb/0JjzIWe29oYYx72PeccY8zvjDHLPLcvMsb83hjzJbDFGJNjjNnHV9HfaIyZZYwp9ax/qjFmga/yv9AYMzWCbb2vMeYT322fGGP29dxWZoy50RjzPrAVUNuLiMRMAVlEMoa19mNgGTABV1EejAuCu+Eqp1cBGGPGA48AlwIdgQOARUEe8nrgdaAT0Ae4I8RT3wF0wIWzicDJwGme2/cG5gJdgT8DDxhjTAQvaSIwDDjEGHMUcAVwNFCMqzj/y/d6CoE3gCeAbsDxwN3GmN0DH9C37gnAfN/3WcB/gVm4bTQJ+LUx5hDfXa4GBvhe28HAiUHGeQLwY9y27A68BNwAdAZ+CzxrjCn2PfftwGG+yv++wBe+xwi6rY0xnX2PdzvQBfgr8JIxpovn+U8CzgaKgMWhNqaISKQUkEUk06zABbOzgYutteuttRXATbjgCHAG8E9r7RvW2lpr7XJr7bdBHmsHrn2jl7V2u7X2vcAVjDHZvse93FpbYa1dBPwFF9r8Fltr77PW1gAPAz1xQTKca6y1W3xV33OAP1pr51hrd/pez2hfFfkIYJG19kFr7U5r7efAs8AUz2P91hizEagA9veMbxxQbK29zlpb7eurvs+zrY4FbrLWbrDWLsMF1UC3W2uX+sZ5IvCytfZl37Z9A5gJHO5btxbYwxjTxlq70lr7tW95qG39Y+A7a+2jvtf2L+Bb4Cee53/IWvu17/YdEWxXEZEmKSCLSKbpDeQAbYFPfR/zbwRexVVeAfoC30fwWL8DDPCxcQfUnR5kna5ALg0rl4t94/Bb5f/CWrvV92W7CJ5/qefr/sBtntez3je23r7b9vbf5rt9KtDDc/9brbUdcdXgbcAQz+P2CrjvFdQH+F4B4/B+HWqcUwIeb3+gp7V2C64N5hxgpTHmJVN/sGCobd2LxlXhwO0bbEwiIs2W0gMqRETiyRgzDhecXsAdsDfcWrs8yKpLgV3DPZ61dhVwlu+x9wfeNMa8Y62d71ltLfXVz298y/rhDoSLlQ0Y843W2scDV/JVkWdYaw8O+4DWLjHGXAQ8bIz5n+9xF1prB4W4y0pcy4P/tfWNYJyPWmvPCvH8rwGvGdcPfgOuWj0h1LbGfSLQP+Bh+uHe8AR7fhGRmKmCLCItnjGmvTHmCOBJ4DFr7Sxc8PqbMaabb53enr7aB4DTjDGTjDFZvtsaTXtmjJlijOnj+3YDLojVetfxtU38G7jRGFPkC6uXAI/F+WXeA1xujBnuG1sHY4y/heJ/wGBjzEnGmFzfZZwxZliwB/K1PazAtaF8DFT4DrRrY4zJNu4gx3G+1f/te95OxpjewPlhxvkY8BNjzCG+xyowxpQaY/oYY7obd3BkIVAFVOLbnk1s65d9r+0XvgMAjwN2971mEZGEUEAWkZbsv8aYClzV8krcAVz+g+N+jzsQ7UNjzGbgTXxtBb6D+U4D/gZsAmbQuEoJrj/3I2NMJfAf4KIQcx9fAGwBFgDv4Q6W+2c8XqCftfZ54E/Ak77XMxs4zHdbBfAjXN/wClxLx5+A/CYe8hZcW0MOrod5NLAQVxG/H3fQIcB1uAMfF+K24TO4cBtqnEsB/wGFa3A/m0tx/2+ycG8eVuBaRCYC5/ruGnRbW2vX+cb3G2Cdb8xHWGvXNvHaRERiYqzVJ1MiIhIZY8y5wPHW2ompHouISKKogiwiIiEZY3oaY/bztaIMwVVyn0/1uEREEimpAdkYc74xZqYxpsoY81CIda4yboL8yckcm4iIBJUH3IubHu4t4EXg7pSOSEQkwZI9i8UK3FHLhwBtAm80xuyKm7dzZZLHJSIiQVhrFwN7pHocIiLJlNQKsrX2OWvtC7gDLYK5C3dgTXXSBiUiIiIi4pE28yD7piuqsta+3NQZWI0xZ+OmJqKwsHCvoUMbzcwkIiIiIq3Ip59+utZaWxx+zcikRUA2xhThTpsayST304BpAGPHjrUzZ85M8OhEREREJJ0ZYwLPuBmTdJnF4hrcmZcWpXgcIiIiItLKpUtAngRcaIxZZYxZhTuV6b+NMb9P8bhEREREpJVJaouFMSbH95zZQLYxpgDYiQvIuZ5VP8GdbemVZI5PRERERCTZFeQ/ANuAy4ATfV//wVq7zlq7yn8BaoAN1trKJI9PRERERFq5pFaQrbXX4PqNw603INFjEREREREJJl16kEVERERE0kJaTPMmIumjqqqK9evXU1FRQU1NTaqHI5JRsrOzKSoqonPnzuTn56d6OCISggKyiNSpqqpiyZIldOrUiQEDBpCbm0tTJ+4RkchZa9mxYwebN29myZIl9OvXTyFZJE2pxUJE6qxfv55OnTrRtWtX8vLyFI5F4sgYQ15eHl27dqVTp06sX78+1UMSkRAUkEWkTkVFBe3bt0/1MEQyXvv27amoqEj1MEQkBAVkEalTU1NDbm5u+BVFJCa5ubnq8RdJYwrIItKA2ipEEk+/ZyLpTQFZRERERMRDAVlERERExEMBWUQy3gsvvMABBxxAt27daNOmDf379+enP/0pr776at06Dz30EMYYFi1alLBxxPs5SktLMcZgjCE7O5tOnToxevRoLrjgAr7++uu4PEcoAwYM4NRTT414/UWLFmGM4aGHHkrYmAL5t01TlwEDBiRtPCLScmgeZBHJaLfffjsXXXQRp59+OpdeeimFhYV8//33vPTSS7z11lsceuihAPz4xz+mvLycnj17pnjE0Rk5ciT33nsvAJs3b2b27Nn885//5J577uG2227jV7/6VUKe9/nnn49qxpOePXtSXl7OrrvumpDxBFNeXt7g+5/97GeMGjWKa665pm6Z5iEWkWAUkEUko91666389Kc/5YEHHqhbdtBBB3HWWWdRW1tbt6y4uJji4uJUDDEmRUVF7LPPPnXf/+hHP+KCCy7ghBNO4IILLmDcuHGMGzcu7s+75557RrV+fn5+g3EmQ+Dz5efn07Vr1ybHUVNTg7WWnBz9exRpzdRiISIZbf369fTo0SPobVlZ9X8Cg7U/DBgwgBNPPJEnn3ySYcOGUVhYyNixY3nvvfcaPdbf//53BgwYQEFBAePHj+eDDz6IuA1h2rRpjBo1ioKCArp27coZZ5wR00kkcnNzufvuu8nJyeH2229vcNusWbM48sgj6dSpE23atGG//fbj3XffbfQYM2bM4OCDD6ZDhw4UFhYyatSoBm8yAl/bqlWrOOWUU+jVqxf5+fn07NmTI444gtWrVwOhWywee+yxBq/9pJNOYuXKlQ3WiebnEC1jDFdeeSU333wzAwcOJC8vj6+++qpuG0yaNImioiIKCws55JBDmD17dqPHeO6559hnn31o27YtHTt2ZMqUKSxZsiTmsYlI6iggi0h4xqTHpRnGjx/Pww8/zC233MK8efOivv+7777LX/7yF66//nqeeuopampqOOKII9i4cWPdOvfffz8XX3wxkydP5sUXX+TUU0/lF7/4RYN1Qrnssss477zzmDx5Mv/5z3+45ZZbePXVVznssMNimie3W7dujB07lvfff79u2Weffca+++7L+vXrue+++3j22Wfp0qULkydP5tNPP61b78UXX2TSpElUV1dz77338uKLL3L66aezePHikM930kknUV5ezi233MIbb7zB7bffTp8+fdi6dWvI+0ybNo2TTjqJYcOG8dxzz3HzzTfz2muvMXHiRCorKxusG8nPobkeeughXnrpJW699VZeeuklevXqxUsvvcSkSZNo164djz32GE888QQVFRVMmDCBpUuX1t33nnvu4ZhjjmH33XfnmWee4d5772X27NlMnDhRJwIRacmstS32stdee1kRiZ9vvvkm+A2QHpdmmDt3rh0xYoQFLGC7dOlijz/+ePvaa681WO/BBx+0gF24cGHdsv79+9uOHTva9evX1y375JNPLGAff/xxa621NTU1tk+fPvawww5r8HjPPvusBewpp5wS8jkWLlxos7Ky7LXXXtvgvu+9954F7PPPP9/ka5s4caLdb7/9Qt5+/PHH24KCgrrvDzroIDt06FBbVVVVt2znzp126NCh9qijjrLWWltbW2v79+9v99prL1tTUxPysfv379/gtRUWFtrbbrst5PoLFy60gH3wwQfrnrdbt262tLS0wXrvvvuuBRo8ViQ/h0j079/fTp06tcEywPbs2dNu3bq1wfJdd93VHnTQQQ2Wbdq0yXbp0sVedNFF1lprKyoqbPv27e1pp53WYL0FCxbY3Nxc+7e//a3J8YT8fRORqAEzbRwzpirIIhJe6qOxuzTD4MGD+fzzz5kxYwZXXnklo0eP5vnnn+eQQw7hhhtuCHv/kpISOnXqVPf9iBEjAOo+Ql+2bBnLli1jypQpDe531FFHhe1jfeONN6itrWXq1Kns3Lmz7rL33ntTVFTEO++8E+3LbcBaW3dCim3btjFjxgymTJlCVlZW3XNZa5k8eXLdc82dO5fFixdz5plnNmhBCWfcuHHccsst3HbbbXz11VfYMD+vuXPnsnr1aqZOndpg+f7770///v2ZMWNGg+Xhfg6xOPTQQ2nTpk3d99999x3ff/99o59L27ZtKSkpqdtW5eXlbN68udF6ffv2ZejQoTH//EQkdRSQRSTjZWdnc8ABB3DDDTfw5ptvsmDBAkaMGMG1117Lhg0bmrxv586dG3zvn/Vg+/btAHX9st26dWv0nF27dm3ysf39ubvtthu5ubkNLhUVFaxbty7yFxnE0qVL62blWL9+PTU1NVx//fWNnuvOO+9kw4YN1NbW1j1nnz59onqup556iiOPPJI///nPjBw5kt69e3Pdddc1OBDSy99jHWzWkB49ejTqwQ73c4hF4Bj8P5czzjij0bb63//+V7eN/OtNnjy50XpfffVVzD8/EUkdHaYrIq1Or169OPPMM7nooov47rvvGD9+fLMfyx+u/GHJr6amhrVr1zZ53y5dugDw+uuvN6iOBt7eHKtXr2bmzJkcf/zxAHTs2JGsrCzOO+88Tj755KD3ycrKqgv1y5cvj+r5unXrxl133cVdd93F3Llzefjhh7n66qspLi7m3HPPbbS+P/CuWrWq0W2rVq1ir732iur5YxF42mf/dv/jH//I5MmTG62fl5fXYL2HHnqI4cOHN1qvqKgo3kMVkSRRQBaRjLZy5cqgVcpvv/0WIOQMF5Hq06cPffr04emnn+a0006rW/7CCy+wc+fOJu978MEHk5WVxZIlSzj44INjGofXjh07+NWvfsXOnTu58MILASgsLGTChAnMmjWLMWPGhGyfGDx4MAMGDOD+++/n7LPPbhQeIzFkyBBuuukm7rnnnqCzPvjX6d69O08++SRnnHFG3fIPPviAxYsX85vf/Cbq542XIUOGMGDAAL7++msuu+yykOvtu+++FBUVMX/+fE455ZQkjlBEEk0BWUQy2h577MHkyZM5/PDDGThwIJs3b+bll1/mnnvu4dhjj6Vfv34xPX5WVhZXX301Z511FmeeeSZTpkxhwYIF3HzzzXTo0KHJPt5dd92V3//+95x//vnMnTuXiRMnUlBQwNKlS3njjTc488wzOfDAA5t8/oqKCj788MO6r7/66isefPBB5s6dy913392gEvvXv/6VAw44gEMOOYQzzjiDnj17snbtWj777DNqamq4+eabMcbw97//naOPPpqDDjqIc845h+LiYubMmcPq1au59tprG41h06ZNTJ48malTpzJ06FByc3N58cUX2bBhAz/60Y+Cjjs7O5vrrruOX/7yl5x44omceOKJLF++nCuvvJJBgwZx+umnR7L5E8IYw1133cVRRx1FdXU1xx57LF27duWHH37ggw8+oF+/flxyySW0b9+eW265hfPOO481a9Zw2GGH0aFDB5YvX86MGTMoLS3lF7/4Rcpeh4g0nwKyiGS0G2+8kZdffpmrrrqKH374gezsbAYPHszNN9/Mr3/967g8x5lnnkllZSV/+9vfeOyxx9hjjz147LHHOPLII+nQoUOT973pppsYNmxYXXuCMYa+ffsyadIkBg0aFPa5v/zyS0pKSjDGUFRUxMCBAyktLeXJJ59s9LH/mDFj+OSTT7j22mu58MIL2bRpE8XFxYwZM4Zzzjmnbr2jjjqKN954g+uvv76uurvrrruG3F4FBQWMGTOG++67j8WLF5OVlcWQIUN4/PHHOeqoo0KO/eyzz6Zt27bccsstHHXUUbRr147DDz+cP//5zxQWFoZ97Yl0+OGH884773DjjTdy5plnsm3bNnr06ME+++zDcccdV7feL3/5S/r27cstt9zCE088wc6dO+nduzcTJkxg9OjRqXsBIhITE+5I43Q2duxYO3PmzFQPQyRjzJkzh2HDhqV6GBlh5syZjBs3jkceeYSTTjop1cORNKTfN5H4McZ8aq0dG6/HUwVZRCRGCxcu5K677mLChAm0b9+eOXPmcNNNNzFw4ECOOeaYVA9PRESipIAsIhKjNm3aMHv2bB555BE2bNhAp06dmDx5MjfffDNt27ZN9fBERCRKCsgiIjHq0aMHr776aqqHISIicaIThYiIiIiIeCggi0gDLfnAXZGWQr9nIulNAVlE6uTl5bFt27ZUD0Mk423btq3udNkikn4UkEWkTteuXVm2bBnr169nx44dqnKJxJG1lh07drB+/XqWLVsW06nERSSxdJCeiNTp0KED+fn5rFmzhnXr1oU9VbKIRCcnJ4eCggL69etHQUFBqocjIiEoIItIAwUFBfTt2zfVwxAREUmZpLZYGGPON8bMNMZUGWMe8izfxxjzhjFmvTFmjTHmaWNMz2SOTUREREQEkt+DvAK4AfhnwPJOwDRgANAfqAAeTOrIRERERERIcouFtfY5AGPMWKCPZ/kr3vWMMXcCM5I5NhERERERSN9ZLA4Avg52gzHmbF+bxsw1a9YkeVgiIiIikunSLiAbY0YCVwGXBrvdWjvNWjvWWju2uLg4uYMTERERkYyXVgHZGLMb8ApwkbX23VSPR0RERERan7SZ5s0Y0x94E7jeWvtoqscjIiIiInGwYAHMn88338An8zow+MTxlOxrUj2qJiU1IBtjcnzPmQ1kG2MKgJ1Ad+At4E5r7T3JHJOIiIiIJMj69TBsGFRXszuwO3Di/U/BjGMpKUn14EJLdovFH4BtwGXAib6v/wCcCewCXGOMqfRfkjw2EREREYmnH36A6mqqc9uygIEADNg5n7Ky1A4rnKQGZGvtNdZaE3C5xlp7re/rdt5LMscmIiIiInG2c6e76rsLj+ScAUCH7EpKS1M4pgik1UF6IiIiIpJBfAG5bfscTjnP1T5P/GllWrdXgAKyiIiIiCTKjh3uOieHgSNcQO5ZlP5dtArIIiIiIpIYvgoyOTnQztc9W6mALCIiIiKtlT8g5+ZCUZH7uqIideOJkAKyiIiIiCSGKsgiIiIiIh6eHmQFZBERERERVZBFRERERDzUgywiIiIi4qEKsoiIiIiIh7cHuW1b9/WWLVBbm7oxRUABWUREREQSw1tBzs6uD8lbt6ZuTBFQQBYRERGRxPD2IEOLabNQQBYRERGRxPBWkKHFHKingCwiIiIiieHtQQZVkEVERESklQusICsgi4iIiEirph5kEREREckE5eXwxz+665i00ApyTqoHICIiIiINlZdDWRmUlkJJSXKf+6v7PqTwl+dwjN2KMbCtD7TpWwzPPAM9e0b3YIE9yC3kID0FZBEREZE08s3tb/LtJU9RUNueY/Mv599vdU1qSK765+OMtbPcNxZYCiz9Dt58E046KboHa6EVZLVYiIiIiKSR4hsv4rSa+7nY/pWfVz1OWVlyn39gm1UAXGz+zsj8uaw9+AR3Q3OqvupBFhEREZFYdaxeU/d19+y1lJYm9/m7VK8EYK8zRnPv24PpOqq3u6E5oTbBFWR/rzQUFcblAX3UYiEiIiKSRnK3ba77+pSfbqJnknuQWeUqyCde2hMGA2/E0DeciB7kL7+Ek09m+5rN9FgJx1m4gkGDm/+AjSkgi4iIiKSLqip38elZuLmJlRPAWljpKsj06OGuYwm1iaggP/88zJpFATCwbmEn0/wHbEwtFiIiIiLpYnNAIN60KbnPX1EBW7dC27b1wdgfauPYgzy/fDWfPrcYFgdctmwJ/5hr1wKw9PSr2D3/ewZlfQ/ucMK4UQVZREREJF0EBuJkB2RfewU9eoDxFWX9QTkOPchzlrZjGLDb3JfhmAGN12/fHubPh+Li0I+5bh0AfQ8azANn7kJZGVxxxXfzoh9caKogi4iIiKSLwApy4PeJ5m+v8M53HEuLRUAP8iub9qWcEhbTj8X0Y1PHftDPd8nPd6939uymH9NXQaZLF0pK4PLLASoiKD1HThVkERGRVKquhmXL+Owz+Ogj2HtvGDMGV0HzBxNpPfwV4549XVhNZQXZLx4tFr6AXHJYRybd+gHV1ZCXB9Nf9pwIZcoUdzIS/xhC8VWQ6do1+vFESAFZREQklcaPh1mzGAOM8S4vKoIFCxIaAiQN+SvGffumJiA3VUGOpcXC14NcUgLTp4c4S6A/lP/wQ9OP6akgJ4oCsoiISKps3AizZlFrslhk+9ct7pe9nJyKCteLqYDcuvgDcd++8PHH6VFBjucsFrhQHPTMgN27u+twAdlfQU5gQFYPsoiISKosXgzA9gFD2aPNAgZnL2CPNgvYMtKXHrZuTeHgJCX8FeTu3V3Vtbq6wbRvCResghxLi0XgPMhNiSQgb9/uZrrIzU1oC1JSA7Ix5nxjzExjTJUx5qGA2yYZY741xmw1xrxtjOkf4mFEREQygy8gtx3an+nT4frr3cfPHXq0dbdv25bCwUlK+CvGHTq4i3dZon38MTzyCABzNgSpIMfjTHpNiSQge6vHJq5THzeQ7AryCuAG4J/ehcaYrsBzwP8BnYGZwFNJHpuIiEhyLVrkrvv3rzsav6QEaNPGLU9QQPafnre8PCEPL7HwV5Dbt3cXSE5A3rCB2v32r/v2lCv71O8fbdu6MLp1K9TURPe4gfMgNyWagJzg1qOkBmRr7XPW2heAdQE3HQ18ba192lq7HbgGGGWMGZrM8YmIiCSVr4JM/4APTdv6KsjxbrEoL2fFCb/hkwmXkH/lb/n1gbMUktNNsApywFRvCXmDs2YNWTtdO8QV3MhnO0ZQVua7LSsLCgvd19FWkaOpIPv7npuaxSIJB+hB+hykNxyY5f/GWrvFGPO9b/m3KRuViIhIIoUKyImqIJ9zDr2+/JILfd+OqvqCsrI3gx8wJanhD8jt2zdusVi5kg0/Pws+3MAP9jgmFVzI9OkhDniLlq/PebbZgz9nXUFenptlok5RkQvHFRX144pEc3qQV692p7wO1kKRiRXkJrQDAj8/2AQ06r42xpzt62OeuWbNmqQMTkREJCHCBeR4V5CXLgXgjuxfA9DHLGsYgiT1/NXiDh0at1j86190+uAlSmo/4Dr7B6qrqa/yxmr7dgAGDMmv64VvELyb24ccTQW5oMC95h07YMOG4OskqYKcLgG5EmgfsKw90OhwSWvtNGvtWGvt2OKmTkMoIiKS7sK1WMSzglxd7UJHVhb7vHg5ALu2W63qcboJVkH2h2bPGebaU0HX3E3xe4PjC8jtuhTU98J7NXcmi2h6kCF8H3ISpniD9AnIXwOj/N8YYwqBXX3LRUREMkp5Odxy7VZYs8YFB++UWpCYCrK/8lZczLjDukJWFjkVG+o/Apf04K0gB7ZYfN0wFr1235L4vcHxTyVXUBD89ubOhRxNBRkiD8gJbrFIag+yMSbH95zZQLYxpgDYCTwP3GKMOQZ4CbgK+NJaq/5jERHJHE89xdLXv+HtR6BdzUYAthf3pSAroF6ViAry6tXuuls3d9BV165u2dq1jQO6pI63guxtsbAWvvnGfT96NHzxBaM6LwVGxOd5fRXksAE52haLaHqQIWRALi937SRnzllLMWTcQXp/AK72fH8icK219hpfOL4TeAz4CDg+yWMTERFJnKVL4fjj6Qtc4Vm8vN0Qdg1cNxEVZG9A9l+vXu0uMQRkf3BpdNpgaZ5gFeTNm93+U1kJxcUwZgx88QUsWRK/5/UH5Pz84LfH2mIRaUD2z2Rx8cVw3XUAbN0GHRbBkRYK8L3mTKogW2uvwU3hFuy2NwFN6yYiIpnJd4ayquLe3LLhLDedbHY2R94QpB6UiAqyvyLnr9D5g3IMB7x/OKOKrANLMRzIpIKb4jejQmtlbX1ALipq2GLhb68YPhz69XNf+w66jItIK8iJ7kEePx7uusv9vvh+Z9oCu3tWqcnOJXuPPaIbR5TSZZo3ERGRzLZxIwD5o3dn0rVX11Vd9wwWKJNRQfYf6O5f3gyzn5vHmfZDurGKP1TfRFmZAnIsPnprC3vX1lJT0Jbs3Ny6gDzvvdW0qf2YvuACct++7g7xDMjp0oN88slwwAHudNI+X3wBZ5zhujVyc+Ge57szrn8GVZBFRERaLf+0VR07UlISJkgmugfZex1DQN57uOtHLaKi8by5EpXycjjpiE3MB9Zsb8/Ccui4vD3DgMFzXoQ5L7oVvRXkVLRYJLoHGWDAgAbfjh4Od+5S38ozLglvwhSQRUREksFXQaZjx/DrJuJEIaEqyDG0WIwY4KqJHbMr1F4Ro7IyaFe9HoD1dKKsDAq2lbCVMfRhGQBZ3Yop/slP6qurKWyxiLj3PNoKcghh31TGmQKyiIhIMkQTkBNxqml/QA7sQY6hguyvJubUVFOyVzWQ1/zHauVKS+GDnJVQDT9k9fRV4zsx4dZPqa6GvDyY/gIU96Fuv6hZvJSP36+lZL84zNobaUB+9FEqX55BwTz4kYUsA5WDod3EsXDPPY3PfhdtD3KaUEAWERFJBn9A7tQp/LrJqCDH4SC9Bv2oFRUJn3ork5WUwN9+txJugD0m96TYVy2dPr1xpbZ8VlsG0YWuNev4ZOJvyX7p/xh/SAT7VVPC9SAPH+6u166l3dq17OlfboG5wNzP4KqroHfvhveLUwU52VrWaEVERFoqTw9yWC3kIL0G/agKyDHbrdDN2lA8sn7avWCtBWVlYBhEV9ZxYc3fmHFrZzjkDw3WKS+HD16r4PAO7zNscI2r7JaUhH6DFq4Hef/9YeFCWLeOL7+Ec8+tP2hueqdjKFi1OHh/cnN6kNNAyxqtiIhIS9WcFosYK8h1faITLSX+ad78wTgeLRaBFWSJjW9as3DzUpeWwnn59/N41TEMZS7D7VcNbi8vh0mT4MHtZzHMPlV/w0EHuZJ0MOFaLMAdPDdgACP3gluH1le2C37VCVYtDv6GThVkERERCak5B+k1p4JcVQVXXsmar1ayeDr0r4VlWTVQUwWFhe4C9UH5hx/4+u4ZfP457Lln/SfpgKs6jh1bH9gDBVaQJTb+gOw/WUYIJSVw59vD+fDRRxj6j73punZug9vLyqC6Gnax37uH7TWGnis+gwULQj9ouBaLIGOoq2z79w/P1Gx11IMsIiIiIUXTgxxLBbmsDP7yF4rxnJK2xnc9eHD9eh07uiO/KisZfl4p3lzcwJFHwosvBr9NFeT4irCCDL6AuvsQ+Acwbx7U1rpTiOOqunl50G6bewOz/oq/0PP8A5ueoi2SCnIo/jddqiCLiIhIVJrbg2xt45kBmuILqpXDxnHB/IvYudNlkyuugEFnHFC/njHwxz+y5I4XWbTIHWtlcJ+i9+uHC0wffwzffBP6uVRBjq8oAjLgTiTSvbs7S+KyZXXzI5eUuE6K3odVwiYYPslXkQ5W4fUL14PclFAV5Npad4G68N5SKCCLiIgkQzQtFtnZrgRYXe0++o6mqlddDUC70btx9gNT6/pEBwWbQ/aSS1hecgmHTqJ+KrEnoF8JsHata8NYty70c3kDcrQnkJCGrI0+IAMMGeIC8ty59ScQwdf+YHw/k+Ji94Zo2zaoqXH7V6BEVJC91eNo3uSlAQVkERGRRLM2uoAMropcXe1CRzShxd9Lmp8f0ckV/NXGRid98LeCbNwYOlSpxSJ+Kircz7pNG2jfPvL7DRkC77wD334LBx9cv9za+jctRUXuTHgVFa7KG+zxo+xBbsAfkAMryC20/xgUkEVERBJv+3YXdvPzIw8gbdvCpk3R9yH7g05e5CftCBqks7NdmN+40V2CTeGmFov48VaPo6m2DhnirmfP5uM3N/Puu7D/xGz23i/HBdS8PF9Dsi8gV1YGD8ixVJBDtVi00P5jUEAWERFJPH//cSQH6Pk192QhngpyzLp0ceF4/frgAVkV5PjxBeSlNT1ZVh7FaZX9AXnaNMZPm8Z43+JlJ19BH3DB2HsdqhUmlh7kUC0WLXQOZICW1TEtIiLSEkXbXgHNP920rwc5LgG5c2d3HaoPWRXkuJk3wwXkjxb3ZNIkN5dxRCZMgBEjqMovYjNFbMW9scp/1TfzSKQBOZYWiwysICsgi4iIJFpzAnKsFeQoWixC8leN168PfrsqyHGz5aUyABbRn+pq1xMekQ4d4Msv+eztzfRos5kfZ70KQKctS93t0VaQE3GQnnqQRUREpJFYAnK0FeR4tliogpw4FRX1LQjLljHq039SQxb/NGeSl+cOmIyG/2DLr5/sBLdDzpbN7oZktFiogiwiIiLR+u5j14O8tiaKHuTmniwkni0WTVWQa2sbBiIF5Mjdf7+r/Hbp4i6jRpFVu5PHOInvsofy979H0YPsUVICZ14asI8VFbnrVFSQ1YMsIiLScnzy8hru+e38yPs8Y1BeDnfftBGA59/qEPlzxlpBjmeLRbAKcmC1UAE5cu++66Zha9sWOnViW5tOfMsQruJarG166umwAj+lUA9ysyggi4hIq1JeDjt/8lPO+csgLj7wi4SH5LIyaLPThccNtR0i7y1tbgU5WS0WgYFYATly/iD54IOwfj1fTF/PmDbfsjy7f7PaKxooLGwYSP3B2F/lVQ9yRBSQRUSkVSkrg5LaDwC4vupSHnkksc9XWgodsl0o2Z5dGHn4SYce5KZaLPxByx9+FJAj5w/IvmDp7x++/np33Zz2ijrGNJxOMJIKsrXqQQ6ggCwiIq3KgftW1X19MG/yyQNfJrSKXFICJx/tQsnJv2oXefjxh45Vq2D5cj79z3L+cvXm8GNN1jRv/kDsPy2yTjUduYCADG4/ufzyGMOxn7fNIpKAvGOHC8k5OcHPmBiOepBFRERatn12XdPg+4N2vh5520Mz9SxyoWTAHu0iv5M/IF9zDfTpw15H9eH864r57YGfNh2SkzXNmz9o+QNyRYULWRJekIAcV9FWkGPpP4aMrCC3vBGLiIjEYk3DgNw7eyXjSxP8nP7g0C6KgHzkkfDcc1BZSUUlZG/eQFu2sXv1F5SV7RW60pisHmR/0OrSxYXx6mr33M0NWfGyfTtfPL+Qjz+G8eNh9OgQ6/XtG93PI578lVZ/sIw3b0COZBaLWPqPISN7kBWQRUSkdQkIyFNLV9A1Hh9rN8UfSqIJZCUlMHcuALPL4YsJF3BuzZ0U5Wxtuo85WT3I/haLdu3Y0aaI3Op1fPJWBeMOT21A3rLHeEZ//xWjw63YvTssXhyf7RStRFeQo22xiKX/GFRBFhERafH8AblbN1i9mq7VKxL/nM0JyB4lJdDv+DbwOFx05lb6NxXo/T3I8WixaN8esrJg82aYNs0dAOb33nsAvP9lEb03FTGAdZx69Gbuf7s4Pn20zbF+PYXff8VOsvmOQQAUF0PXLgHrzZ8PP/wAq1e7SnKyZVqLRQb2ILe8EYuIiMTCH5BHjoQ334SVKxP/nDEGZIDeu7kqXf/iMLNaxLOCnJXlKq0rV8Ivfxl0lQ++7cxBdGYAiyjasZ6ysl1TF5C/+w6Ar80I9sr6nLw8mP4ijT8hGDzYrRvtDCHxkm4BOdYWi7w8t69UV7uqsT8Qq4IsIiISJ1u3urO0AR99BG9/2IaJB2XHL3T5A/KoUS4gr1jhDi7zVkfjzR9KYglE/o+xw4W6eAZkgPvugxdeaLT4iy/gnU/bcoc9nxF8CUDPnDWxzeEbK19A7n3gIK6f7KbYC7rfRLotE6Gmpv5n5J/KL96a22LR3IBsjNu3KyrcNm3f3i1XD7KIiEgcXH89XHVV3bd7A33oxb4Fn/PkW93iE5L9AXnXXV1Q2rLF/WP3/1NPhDhUkP2h7tN3t1Jd3sR0YPFssQD48Y/dJcC2crhsknu69bYr1MItv1vDbqmqHkNdQO66zyAuv7yJ9UK1BCSD9wC9RL0pa+5BerG8qWrb1v0ebdnSOCC3wAqypnkTEZH0MX26uy4ooDqvkGpy6c0KflX1t/hNxeYPyMXF0KuX+zrRbRZxCMjzV7qA/OVHWznwQEJP9RbvCnII3pNblE4pBmC3jmsT+pxh+QIygwc3vV6og8qSIdHtFRB9BTnWHmQI/qZDPcgiIiJx4P/nWlbGp7V7c9mBHzKjqoRz7V088+YezKvOCp192raFww4LXzkNDMjz57s2iyFD4vYyGolDQJ7xcVt2A9qylaoqeOSREFXkJAVkcM9fUgLcVAxP0WiGkKSbN89dDxrU9HqpbLFIRkBOdg8yBH/T0YIryGk1YmPMAOBuoASoAp4Bfm2t3ZnKcYlIy/bVAx/z6acw5KTxqTt4SCLj+fi5ZATc/PY+fDvlIIYuf4vT3zoR3gpz/xtvhCuuaHodb0D2n+RiRQJnsvAfuJSbG1PbQ3WOCyBtSXIPciS6dnXXqQzI1tZXkNM5IPufM9kB2f98W7Y07rmPR4uF9/H91IMcN3cDq4GeQEfgDeBXwO0pHJOItGDl79ey+5kH0xdD7wfX8+ZbWQrJ6SzgBAolJXDvlLv46LabybNVGAMjR8Duuwfcb9MmeOUVeOwxd77epno7k91iEY/+Y2DiYW3hVReQ8/Lg5JNDrBjvHuRIFLsWC9YmqcXCWvj8c9i4EYCvv4Y5H1fw882boUOH+sAeSip7kJPcYvHh7HbscwCuiltQ4MLw9u11BwiWl8Pqf1dxFMS/xUIV5LgZCNxprd0OrDLGvAoMT/GYRKQFe/+NrZSwGYCs6u2UlbVVQE5nQc4wNvLYoUy69yGqq13mm34P7nNGrx07XDV4zhyXlvbYo9FDl5fDjOk7uWzDBhegO3dm0Y5eDABWzFxBr0S9pjgF5N3Hum0ypN82yp5s4iC9VFSQ/QE5WRXk55+HY46p+3Y49WGhstdg2oU7+C0depATdRY94JP5nRgHVJPLQYfmMX26b39p186F4wsvhIICVq6CL56HwTVzAFi9uYBuzX3SYNtUPchx83fgeGNMGdAJOAz4P+8KxpizgbMB+vXrl+ThiUhLUzqu/o91p7wtlJYm7p+SxEGQgOw/GKysrIlpu3Jz4eij3ZRkl1wCY8c2uHnZcnjvCWhTsw2AHR26MPPjbKbd25MHgY5P3UP19KciK7p27Qr//S9E+j8oTgHZv036dNpKn6be5KWyxSJZFeQv3bRyDBzIIjOAhQvAArVks2HcRUwJd/8M70F+89NOfG1OZY3tSnW1+90pKcF9YrJ2Ldx/P+A+rj/Xc7/5ld2bH5D9r2f6dNi2jXnzYNuLHzMKFJDj4B1c+N0MZAMPAy94V7DWTgOmAYwdO9YmeXwi0sKMHVYfkF94fAtjSopTOJo0tHhxo+kQ5s2Dj5b3YbdT909utd3aoAEZPAeDNeWEE1xAfuMNd/HoA1zq+X5tYX/KyuCDmr3ZShvasg3WbotsnCtWwGuvwVlnRbZ+POZAhshCXU2Nu0ByQ0myK8g//OCuL76YlWMv4Me+6eby8mD6ORHcP8N7kEsPNEwqeLBum9TNTf3vf8Prr9ett3Ah3HEn1OyEnTkFnHJF2LcWoXXo4K5vvx1uvx3vsbQrNrZN3Cc0CZI2AdkYkwW8igu/+wLtgH8CfwJ+l8KhiUhL5vm4b8yQFHycmu4OPRS+/bbBosG+y/iHZ3Hb2yObDqbLl9f1gXp98QV88gmMGwejR/sW7rZb01XNHTtcuMvNbd5BPaWl8OijsGRJo5uWLIEHHnAPn5VtOPrmn1C6K1yfvxs9q1ZTnLeJp56CvfYK8xx/+Qv87W+wdGnk44pzBbnJUOfvP87PT+yJTwJ16uTOpLZxo/s5JvqgLH9A7t49sk8YAmV4BTnkNhkypMFsLQOBKVPq1xsfyxviCy+s62/+Zg7Mnu3e824zbdne42wied+STtImIAOdgX64HuQqoMoY8yBwAwrIItJc3n64VJ1WNl3V1NRPi3XccQB88w0UffU+fVlG7x0LKStrIiBPnw6TJwe9abTv0sC4cfDxx6HHE2tvpjFw4olBb+oHHHpKfRAY7XtNLkS0o7S0HXtFEg6GDnXXy5ZFPq5kBuRUtFeAC8ddurgK8rp10KNHYp/PH5B9zxPRJwxewWZcSJZkHKRH5Nsk6m0XyvDh8PDDAGwqh1O9Vf2fx+HxkyxtArK1dq0xZiFwrjHmVlwF+RTwnb9SRKQ5vP8AU/HPMJ2tXetO6dylCzz5JOD+sX024RROrHmE4uwNTZ82+PPP3XWXLtC9e93iNWth9er61boXW7qumePKyk0J0V4RL8GCQNThoE8fd52uFWR/QE7mDBZ+xcUuIK9Zk7yA7NnvopIOFeQEHqSXas2q6qeZtAnIPkfjDtT7PVCDm/Hy4lQOSERaOO8/QAXkhoKEjJISGHB0R3gaLv/VRgY29Y9t/Xp3ffHFcOWVdYvnl8Mkb/XoRehamu8WVFWFrm4mOCDHRd++7jqaCrJ/v4s1IPvbJqqqXPU/O7vxOt4Wi2RLZh9ySw7IyZgHOQ3ErTKdImkVkK21XwClKR6GiGQSVZBDCxEyeu7uTjIwsOPGpu/vD8idOzdYHLR61L69q1hv3lwfpAK1hIDsrSAHnmwhlHhVkI1x22bLFti2LfjjparFAupmslj28HTef7SakSNh2LAw9xkyBAYOjO55tm512zQvr/7AsGilQwU5wwNyS5dWAVlEJO7UgxxaqCqc/yQDGzY0ff8QARmCVI+KijIjIHfs6ILNli3u5CSeEzKEFK+ADPUBeevW9AvIvv2ozyM3cVyk92nf3h3oGc228e63zT0QMdNPFCIxU0AWkcymCnJo4QJykNkpGmgiIDfSvr273rw59DotISAb46rIc+e6NotoAnI8AlG4ymcqe5DPOYe5ZStY9M02LGBwZ3zeZZcQ63/yiduHPvsMDjgg8ueJtb0CMv5EIRI7BWQRyWwKyKGla0BO98pa374uIC9dGvSMfY3Eu4IMoQNyKnuQR4xg/f3P8zNv//nDsEuoPtRzzoF774WZM1MXkNWDLCEoIItIZlNADi1U0OjkepBVQQ7B34cc6YF6yQzIqWyxIMrZC/yTTs+cGd2TtPSArBaLFkEBWUTiqry8mVP7rFsX8T+rTz+FGZ+3p+TQDuGfQz3IoSWwB7mRTArI/pks3n/fnfwkjM1fLqI9MHd5O4aEXTuMdG6x8Il49gL/6cA//TS6J1BAliRQQBaR+Jg2je3X/ZkeK2o5zndw//aeUBBJIWvLloYT54axFzCCXA760wfc8vbYpv8Zq4IcWiwtFtXVUFHhphrzh9+mRBKQW0pvpj8gP/xw3YkRmuLfOr+9rj1XHBTj1Ffp3GIRreHD3TjnzXMHPIaZkcL/5vu0WT/QA2ILyK3gRCESGwVkEYmPBx6gYPn31E3YZIEVUdy/XbuIDnjaXAHZm9ZTyFaGV39OWZkCcrPFEpD91eVOnSKbScAfkCsqQq/TUirIP/sZvP66m5UjjMWLYeFCWEZv3t45gX3LEhyQU9xiEZW8PBg50h2sN3Rok6enrqqGPqthqoUi1rmFsQTkNm3c9datkU/XF8zy5fDWW+4xgvhuPnw7B4YOg0H+DxtWrXLX6b6ft3IKyCISH77K1c/z/sOsncPJzYXHHoMxYyK4b24u9O7tTlcbxtfl8PmEC/lVzR0U5Wxt+kxvkPCAXP6B5dMXl7Hv3jWRvVZwPaw5Kf7zW1tbX7Xv1q3hbUVF7mdRUQE7dwYfazTtFZBZLRZdu8LTT0e06opyONxz0FrY/TWcFtBiEZWjjnIB2R8aQ8gH+nq+35mTT46/RaM5cnLcNvKfvKagoHmPc9xxrtUmhEG+S1CRzIAiKaOALCLxsWMHAP/3wABeXroLpaUwJgFnUSopgX4ntIXH4MLTt9AvhT3I5eUw84BLuKDm79Hdcb/94N13m1+1CsXa+gPCAnz0Ebz3Huy/P+y9N67nu6bGVYADq41ZWe7j7g0b3EffXbo0fsBoA3JRkbvOhIAchbifcjeTKsjgzsB4+ul1fz9C+fRTOOEEt1puLjz2v46M3y2C1p6mtG3rAvLWrc0LyLW1boo6gKlTG73B/2q2O7u6v0A9ejSM8E96sueeiT8dt8REAVlE4sNXQR41Lo9RJyb2qXoPdr17/bpGEHgTWEEuK4OJNR8DsJIetOmYT8dwJ/ZascJVnGbNcv8x42nKFHj22aA37e27NBLqY+qOHV1A3rgxPgE5kyrIUYrrKXczqQfZr2fPsKvs1Q8efrv+jcb4eGzPwkK3f2/ZEvl+7LVsmTujYbdu7uOyAJXl8EvvlHd3AS341MutjQKyiMSHvwLURB9h3ERzgE0CA3JpKRSaSrDws/xX+NvLo8MHofPOg7vvhkcfjSogh50d5K23XDjOymoUMKuq63MTuH/W+Xm4dadODf6E4fqQExmQdfBSaJlWQY5CXN9oQOwzWcyd664HDw56c9w/PZCkUkBuaSoq4IknWDS7ku+/hx6H7cnwCw5K9ahEkhuQozkLVgIDckkJbO9ZASvgH4+2Y89I/gGedJILyI884vp7I7ByJXz2HBTWwmdZMODoIEW3119319deC3/4Q4ObPiuHSd5K1vQI/ln750IONdWbPyAHqy4H04oryHGVaT3IqRRrQJ43z10PCT15X9xDvSRNxAHZGPMCcD/wsrW2NmEjkqbdey9ceikDgAFA9Su5fLT7evaeFIcJ6EVikYoKciT/2BI8D3LBDtfzu+cBRZHdYe+93RH7334Lt98e0V16Auf5v6kBQh0fVlwMF13UaHGzKlmprCArIIfmCXVBP1VoiS0WqRKvCnITAVlarmgqyFuAp4BNxpiHgAettd8lZFQSmu+o83eZwGg+p4hKPnl9gwKypF4rrCAD0Z8lzRj4z3/glVdCTg0VaOFCuOtuqNkJ2Tlw3q9g4MAgK06aVH8wXICoK1megBw0iCUiILeUeZBTybdt3n21khduv41htbM5Kv+vvPhWkfvZZHCLRdxF80Y7GAXkjBZxQLbWTjXGtAemAqcBlxlj3sNVlZ+21m5L0BjFa/t2AF7I/Tndd6yiiO/Yb0+dHUzSgL9ylYyPdiP9x2ZtYgPyzp3uIB1jogt1gwa5S4QGAsccVx9SBybjI1tfQH7zjjksn30ax9eWuZO/9PCd/GXNGrdetAE5E+ZBTqHvVrRlEDBhzn1M8C37oaoHZWXXNwzIarEIL4I32k32/ofpQZaWLaoeZGvtZuAfwD+MMcOBM4F7gduNMU8Bf7fWzon/MKWOLyD/6uIC2j3aFlbCnkMUkCUNpONBelVVbiqmnBw3pdn27e46Ozs+4/A/f7t28Z+yLUCyexmXVHSiHzD5y7/WL7TASs9KubluuqpIqMUiLt5dvwcDySaHGtbSha6s4yL7N/719ki+NbkMneP7F6wKcnj+/ezCC+GqqxrdvGUrtFsAP/ZN07ZlFyj07pqLF7u/JbvskpzxSlI16yA9Y0wv4CjgCGAn8CxuDu8vjTGXW2tvjd8Q01fYo8oTwReQdx1eAANcQE7JueRFAqWixSLcvu8NsDt2uO+3bYu8HSIcfzU0Xo+XRt7OOZjDuY0ObGI6k7iSm9iW35HHH/ec/KVTp8hPduB/U1NZGfpNigJyWMOmjmGXB1bSrno9y7L7M73zFMb98D/OeuNYeMOzomYCCW/ECHjySVi61F0CFAIj/N9Y4Psgj1FSomp9hormIL1cXCg+HTgY+Bz4M/Ava22lb50jgUeAzAnI//kPXHddo6PNt2yFNvOzmM2lXF9wQmRHhceDLyBTUBD7AQYi8VJTUz8bfryqs02JtILsv72w0LWAbNniLvEKtP7+4xB9vy3Z4JP2pv9Da6iudgX4006Dk0+O4eQvWVluO1VUuO3WIciE0QrIYZWUwFNvFVNWVkxpKXzy/B3Mv7WINnYrBhg2DAbv0xmOPjrVQ01/l18OxxxT35YSYNYsOOus+pOT3HcfjBoVsJL6jzNWNBXklYABngAus9Z+GWSdd4AQcwK1UNOmuVP4BCgERgPnchdPVZ9AWVkKAnKsBxiIxEsy+48h8oP0vAE5N9f1zcazDznaA/RakITM4dq+vQvImzc3HZBV/WxSw3abAUy684n6KfweQCejiJQxTQbcUSPhtt3qfwdGabu2KtEE5ItxB+NtD7WCtXYj7niSzOH/x3/33Q3+Q3z3/GwGXXcS+VSRl+d+eZJCFWRJR8lsr4DI3xwGBmTvsnjI4BYLSEDfc/v2sHy5mwO6Vy/mz4dv58LQIbDbbtS/4VAFOWI6GUViaR7j1iuaWSweTeRA0pa/tWLIkAZnvRoEcB3061HN9OeS34OsgCxpJVUBOcIK8uaatlRuyqUX8PDdWxh8cpx+ZzO4xSIh/CcV8Z3IZDffpYH8fPV0RkkhTiT+dCa9cEL94/cdIdy9YzXdk/mHKVhATsTcriLRSHZAzs93H49WV7s3sTkh/pT5fjc+ml1ITm01vYDd77mAjdM6892VJzLoupNiG0cGt1gkxM03wz//CbW1fPklfPZ5fev6mD1h5Ejg4IMTPiOIiEg4Csjh+CvIgf+A/RUOfwtGsqiCLOnIH5CTVfkzxlWRKyvd/u+fQizQ2rXuynahkkIOpIxxzIRa2HjHXIg1IGd4i0Xc7befuwBbyuFX3tNf34l6Z0UkbWSlegBpL1RlTAFZpJ7/9yBZFWSIbP//4QcA1mZ15yJzB5N4kynmGQDa1TQxH2+k1GLRbP7e2euvJ3mzAImIREgV5HBUQRYJL9ktFhBZH7IvIB95Vncq+xXQpcskNv5QBVdBzvbK2MegFouYqHdWRNKVAnI44SrIIeZPTBgFZElHaR6Q+4/vzuWn+ZbZPLgux43Z//l+c6nFQkQkIykgh5POFWTNgyzBVFfD66/XVTdnVQ/j5eWjEjsFVLJ7kCGqFgu6d69fZowLtBs3unAdy5jVYiEikpEUkMPx/+NPx4CsCrIEM20aXHBB3bdDyOfQrCVcn98tcb2eqehBjqKC3CAg+++7caMLuJ06NX8MarEQEclIOkgvHH8FOfAff06Oq0TV1LhLMtTUuMBujBuPpnmTYJYtc9cjR7Kx0wAKqOLg2leprnYnE0iIVLRYNLeCDPWBNtbfHbVYiIhkJAXkcEJVkI2pryL710k0f79zQYF7flWQJRj/pwynncaGUy8B4AjzUmLP+JiOPcg1Ne7U0gDdugW/b2WMB+qpxUJEJCMpIIcTqoIMyW+z8LZXQIsJyOXl8Mc/umtJAs9+MvC8wwE4Ivc1TjtpZ+KeMxU9yOF68Netg9pa10IROC5VkEVEpAlp14NsjDkeuBroB6wCTrXWvpuyAYWqIIMCcqCPPnL9r56Wk9VrYP5r0L02m9/n/ZI/vT1e0zolmn8/yc+HXXdlW78htF0yl6un9aJ2WjbVXSBvQG/3sxozhvJy13oR00F8qWyxCBVyV69214HtFaAKsoiINCmtArIx5mDgT8BxwMdAz9SOiMgqyMma6i3dA/JVV7nZEzy6Af5zlfWsWklZ2csKyInmbcUBPhx2GgcuuYxu+NoN1gHrVsGECVT2H07utzDZd7rfyqGeYmiPHvDoo9ChQ/jnTMeD9EL1H0P8Ksg6SE9EJCOlVUAGrgWus9Z+6Pt+eSoHA6R3BTndpnnz93teey306wfA/PnwzJ++57KdN9DZbEhcD6zUC9hPCq7+PX1nnIGt3kFeHjz9VC17PXsFPPII7eZ8wlj//SwwJ+CxXn0Vjjsu/HOm40F6kQTkWCvIarEQEclIaROQjTHZwFjgP8aY+UAB8AJwqbV2m2e9s4GzAfr5QlhCqQc5chs3uuupU2HXXQHYDThs6Fdw0g2MGFhJW1WPEy9gPykpgX+/1bWujWKvEuAnD8Hll/PV+5s57zyXb3Nz4a67YMQI4Lbb4IknYOnSyJ4zzQ7SKy+HLY//wGSIvcWiqgpeeaXRut99B7tsqiAb1GIhIpJh0iYgA92BXODnwARgB/Ai8AfgSv9K1tppwDSAsWPH2oSOqLbWXQCyghzPmJ/vrhWQHX9A7tixweJR+7ow0rY2Dqf2lfD8LRb+/ZMgp/Q1BoYOZcRQ+NPu9T3II/zrjBnTvICcxIP0FvxQyC7AqoVb6QHw4ovwpz+xZc0WCr+HXewqAJZUdafRW+loWiymTYMLL2y0eJDveittmPVpHiX7Nu91iIhI+kmngOyvEt9hrV0JYIz5KwEBOam81WNjGt+e6gpybq5r/di5M/ZT5sbKWti0yX3dvn3D24J8nB30wLCtW+tfI/DJJ/Dee7D//jBunG+hMS6AB/t5iBO4n4TRKDwD9O3rriMNyEnuQS4vh3/e1pb7gKLnH6Gq23/JX+M6sgqBkd51q/dqHJCjqSB/9527HjsWhgwBYPbXMGuW2+1fN4cybIZRQBYRySBpE5CttRuMMctwnZB1i1M1HqDp/mNIfUAGV0XevNmFy1QG5MpKV20vLGwckgIC8udPz6fdcUdzkl2PAaqKId9Wwdq1De42zndp5Iwz4P774/0KMkeUATkof0BesiSy9ZPcYlFWBp/WjKaKPArZCmu2uor5jTcyq3gyZ5/thrQltyMPnd2/8QNEU0H2z4bx61+79iGgohzOmlT/vnR6aTxelYiIpIu0Ccg+DwIXGGNexbVYXAz8L2Wjaar/GNIvIAe0NiRViPYKANq0cRXf7duhpobVj77GIfar+tt9x/aRl1dX2du2HbZta/gQbfJqXZX6rbcS8Qoyh3eat+by9/enaQ9yaSlcnz+KHlVr6Jq7iSefhL0mdYSiIkYBfx8UZuq6aCrI/oDsOdlISQlMnx6H6fFERCQtpVtAvh7oCswDtgP/Bm5M2WgirSCnapo3SJ8+ZH97RbApwYxxFbuKCtiyhWG93Lp3mfP4e/5lLtyMz3YHU/l6vb8oh0neCt10KNl9kwvgAZVmCRAwzVuz9Ojh9vvVq93jhQvbSe5Brg+o7Sktbe8OPAy4vcnQGk0F2T87S3FxdM8hIiItVloFZGvtDuBXvkvcNPtECC2lggypD8hNVZChPiBXVtKvgwvIu/+oD49c3adRuIEQFTrb3oW2iorIQltrFY8Wi+xs6NXLtVgsW1Y3K0lIKZjFIqaAGs00b0EqyCIiktnSKiDHRU0N7LsvfPxx3aLxZFFmLmNSwY2uEhnpP1V/QE7nHuR0mQs5XED2fqTtqzYfeFQHaOJnEXTmha5dYdUqdxrhXr1iG3N1NXz1Vf1MJQG+/BI++8xN6DByZNBV6nXuHD5EJks8AjK4PuQlS9wl3GtLxYlCYhFpi0Vtbf0nFl27JnZMIiKSNjIvIK9c2SAcA2RTy0/tc/xf9Y2UlUURkMNVxVI9zRukXwU51FnXvBW7ULNdRMIfkNeujT0g/+IX8OyzIW8eScPZEMIqK4OJE2MbUzzEowcZXB/y++9H1oecinmQYxFpi8X69S4kd+yY2oNgRUQkqTIvIPuD4q67wnff8cmr6xh3eDE9WEVeHtGdya0FVJA3VLWlEzDns60MOyg5wwjKH3qbarEAF0ia6lcOx1/Fi0cf8gcfuOvRoxv9jFeuhGWe8zj26Q09Q534fMMG+P57+Mtf0iMgx6MHGepnsgiYD3nJEli4EIoPH8fuv/6RW5iCeZBjEmkFWe0VIiKtUuYG5MJCMIZxh3SmNieXTjs38tbL29mnJIrQEK4qluKAXF4OKz5qyzFAwe8upPL+q2lXiAs2TzxRX11OhuZUkFMZkLdscSk4Lw9mznQ9tx6LAg8SfBp6hvrkYfVqt83/9z94++1GB3MF+uILN8fzuHEum9fp3Rs6dYrlVbk3dTU17vWEemMXqd12c9evveYuPv18l6o38vho5Hr2PqgwcyvICsgiIq1S5gZkfzjMyiKrR3dYtox9Bv4ABJkTNZQ0ryCXlUF17XCO4WkG2oUwd6G7/bPPYMYMOOyw5IwLIjtID1xA3rzZfZ3KgPz99+564MBG4RiinMarWzc4/nh45BE4KHwZf7Tv0kinTrBgQWzT9cWr/xhcC8qmTa5C7vP+B/DODDjP3kF7Kvjk9Q0uIGdqD3KIGSxERCSzZX5ABjdl1bJlrne1fxQBOdIKcoqmeSsthUn5V/Fc9TG0y63iH/+AkS9e7065u2xZ5I+7bRuzH/mMzz+zjN4rmxGnj4u++hiuxSLIQXopDcjz57trf5U0iKhmSfjDH2DePDfDRhPWrIEfVtd/372bL3stW+aC6BtvwJQpET5pEEFOM91shYXw2982WJRVDtdPgqO3PUN7Kth/lO/1qoIsIiIZJPMCsv/sEoEBGVxAjkaaV5BLSmD6W4aysj0oLYWRJcDCkVEH5A2TjmGP8lfYw/d9xR0lFM18m/LP8iOfHi+ZB+lB/CrITQTkqAwa5Hpewpgf2LrxAhSXAH/6E1x2Gbz6amwBOZ4V5CD8lfX2U9rDchi9a0BAbik9yPn5bs7tqio39lDBXgFZRKRVykr1AOIuVAUZog/ILeBU0yUlcPnlngDrP7Aq0jOgAXbedwB8yhjW0JWi2eVsnHgUrx5wE9f/oYpJkyLIfpG2WFRU1FdZUxmQI6ggJ4I/YF5/PQ2nHDzkEHf92mtg68+wXl4Of/xjRNnbSXBABjfmnoOL3Df+dpmWVkH2n7wGmq4i+1ssFJBFRFqVzKsgNxWQV66M7rHCnSgkHaZ5C9Snj7uOooJcmOM+lv951vP0zl3NO7X70fGj17iW15jLLjxTfXz46fEincVi5UoXANu1C9r7G1a8A3IK5i4O2roxcqQ7k+Dy5XDeedCuHctXQPlTkFsD5dnQ7zjo3QtXpT///ODV+nhN8RZOkS8gV7TQFgtw++Dmza7vu0cPZs50b0RKSmDsWN86S5a4a/Ugi4i0Kq0rIGdgBbkRf0COooKcX+se99eXFTD+iLFkFX7Czn0nkLNlMyPNbP4TyfR4kbZYLF/e9HrhRBGQmzyDYooqyCFlZcHhh8ODD8I//gFAb+AS/+07gcc96/tDcqB4TfEWTmBAbmkH6UF9X/xeewEw1ncJSgFZRKRVydyA3KZN/bJYe5DTdJq3oLwtFta6j5IjfNyLfl8A7QFGkjPtHzB1KkfvPocD74uiBzncQXorVrjrRAXkzZuhtJSaOXMZub3+RB81BZDtbSjautVVsKM5aDPR/vQnGDWqLuQuXgzTptXP2nb22dD/+7dcG0aofTkJLRZAfXtMS64gn3463HEHWEtFZcPjK4uKoMj3no6BA2HvvVMyRBERSY3MDcjeCrL/DA/pVkFetcqlnwh8+in0nLOJXtB0+OnQwYVR/wk5IpkyLFioGjrUXdlv608HvWkTPPAAS7+pYNEiGDCgPo+zfr27DtdikaCA7K8Wn/TdrfT5/HOygULvCtuDPNYRR6TXQWXFxXDRRXXf9geOOLG+Ct6/BLijjQvInqnXGkhWQC4K0YOcTtsznMsucxdgduDBk69FMYOJiIhknNYRkBNdQW7ONG+/+x3cckvEq+/l+fqr7woYcUCIFY1xbRZz57o+5HABubY2ePVvyBB3/d13bjvk5Lhy5u9+R1+gb6MHwoXeUMEssMWiOQfogfu5FhS42Up+8xvIzq7r1c2rsXS0rj3h6zvfZuJvx9UFnpdeClIETOaJVJqpUb+y/0Qi/op9IPUgN0tU816LiEjGax0BuXt3AHYuX8UnH1hK9o2g7QASW0F+91133bVr2KpbhefcGgvYlfLlIxnR1B369nUBeelS2GOPptZs2LPqbccoLIR+/erPKzxoUN1Bjq/zI8rZhywDBx4I++/vu09paeiWDn9Arq11182tIBvj2iLmzoW//hUI6NUF5g07kuHnlfLfMfWBZ+9MCTz+gByqgpysHuTAFouW2IMcIKp5r0VEJKO1ioBc/mUhwymi/c4Kfj5pA8+81Tmyf4SJ7EH2tyS88w4MG9bkqo0+/j04zGP7D9SbOZMvNvTn449h/HgYPTbHBV1viG3qI/mhQ11A/vZbdz9f1fL53GO5r/YM8vJg8g3Ut2A0xR+Q/ZobkAGeeqrBqY+9vbo1OXkce+tUIEMDj/8TgXRtsWjBAVlERMSvVQTksjIoYFf25AsGVM+jrGyfyIJTuApyLNO8+QNyly5hV436419/Y/BVVzGaqxqe1vjii+sqr0CTgWplh6H05HUWvzqH/j/5SV1A/vU1HelnovwourCw4fexBORRo9zFJ7BXd1ymhWKvSFsskj2LRUvsQRYREQmhVQTk0lKYlz2MPWu+YI/sOZSW7hPZYyWqglxbWx+Q/YEnjKiqoccdB6+/ztoFm1ntO89BIVvozxL48suG64YIVOXl8OQLQ7kNKLt3LoNPhBJfKBuyd0cunxThWPziWUEOIiOrxcGEqyDH81TTTcmEWSxERERCyLwz6QU51XRJCZSc7toYrjp2TuRBKlE9yBUVLiQXFSUmUAwfDh9+yHcvfsPYNt8wMvsbTst7wt3m3z5+IQJyWRksqekFQJea1ZSVUX8ykOaE24CAvHB9fANyq5FuFWS1WIiISAbKvIAcbB5koN+PXEDuvfGbBsubPJVvoirI/upx587R3S9K3tMa/+1e3xsG//bxCzHrQWkpbMpxU6oVm7XuRCHh5jpuSkBAvunO9pGfPlnqFRS4/a6qqvGbHUhdi0UGHKQnIiLi16JbLLat28rtl69k3JE966vCwWaxgPoD4ebMqVtUHnjw2/SAj+kjrSBHO83bunXuOsEBGTytB3NDBOQQsx6UlMBfHuoCv4CRfdbRpoTYAnJeHlvbdqHtVvfaF9X0DX/6amnMGFdF/uEH12YR8EYwadO8hWqxUA+yiIhkgBZdQW6zaA7z//QskyZ5KsChAvKgQe50ZAsX1lXeyspcOK6pcddlZQFPkOgKcgQH6MVN2zAV5CAVxz0PdhXkNpVr3Vn5YmmxAObf+Rq/yb2N481TfJBXGv701RKc/w1KsDaLZJ9qWi0WIiKSgVp0QAboblc2DLehAnJeHhu77grWMuOXj8POnZSWuoybne2uGwW2RPUgJ6nFooFmBGQ6dXIVyw0bXDiuqXGP08wQNPK0vfj5jAsZdeOxvPlWlqrHzdXUXMjJbrGorHRvnhSQRUQkg7ToFguAnmZVw3AbIiBPmwbFP+zOz5jHxEfP4qsVX1Py5t+anj4tXAW5udO8tZSAnJ3twtj69a7yDs1rr/BoNbNNJFJTFeRkBeScHNfesW2b26fUgywiIhmkxQfkAwavYvqDhO1BfvZZ2MbFHMqrtGE7OTNdT0aTgS2TKsj+wLR9u5tBIyur/nsI3bPatasb7/z57vsYA7LEQSQV5ET3IIOrIm/b5tos1IMsIiIZpMW3WOzWdmV9wK2trT+yP+DgpWOOgXc5gAEsAmCXHXPdR8NNaeGzWDRgTP2bBu/sB+F6Vru6PmS+/95dKyCnXlMBOVk9yNBwJgu1WIiISAZp8RVkVq2q/9r78XJWw+x/9tnu+tlnulH9bhH5Wze62ST8ATCYTKoggwvIW7e6i//MduE+kvcfSOgPyHE+wYc0Qzq0WEB9QN60yb3ZzMpq9HsnIiLSErX8/2arV7uDxyD0AXo+Z58Nr71uyBs+2C347rumHzvSCnK007ylYhYLCN6HHC5Q+d9AqMUifaTDQXrAZtxUb3Pe801bqOqxiIhkiJYdkLOzXTj2zysc4iQhjQwa5K7nzWt6vURVkJM4D3IDsQRktVikj4AKcoOT3STpVNPl5fD+l66CvOG3N7iF6j8WEZEM0bJbLHJzXUBeuRK6dQtbQa4z2FdBDheQ/RXkSAKyta7PNxKpbLGA6AKyv8q9dKm7VkBOPX8F+Z13WDnlAr54Hopq4Yss2KvTHPIg4RXksjJobwcAsG/t+25hz54JfU4REZFkafkBeft214c8alT9wWeRBuRwLRbhDjzKzq6rYt98Yw0TJ+VENoVZOgbkpmax8FIPcur16+euv/+ent/fybn+5TXAWt/X3boldAilpfDT/Jt4p3oibbOrueIKGHTKvgl9ThERkWRJy4BsjBkEfAU8Y609MeSK/uDqP1Av0gqyv8Xi6ad55qQX6H3uUZTsG6T6G66CDNTk5JFds43T/q83/J+hujPkhWvFXL3aXSc7IPtbT7wBOdJZLPxUQU69sWPhuedg2TIWLoQ77oSanZCdAxecDwMn7VK/jydISQm88FZ7ysqmUFoKgzS3tYiIZJC0DMjAXcAnYddqbkAePBiblYWpreXnj/2M/Z6eya1v79W4+hvB1FVLe+/DgAVv0x1f6F0fdtTO3nsnv2czlhYLPwXk1DMGfvYzAAYCU6bUn+xmYBKDqk76IiIimSrtArIx5nhgI/ABsFuTK/uC6yf/WcnOA6Ak0oDcsSP//fkjHPlvV5zusWMJZWVBAnIEFeSVj7zJgZNXU13t8u7TT7sCX1jFxRGsFGfNCMifL+3Knt4FCshpR0FVREQkvtIqIBtj2gPXAQcBZ4Zbv6rGBeTCD17nzQmXMHT8XDpB+IAMFP96Ko8++wYn1TxMcfaG+lNVe0VQQS7ZL4sn3upRV8Ebm85BJcqAXF4OR5/eg6Vkk4NvKj0diCUiIiIZLq0CMnA98IC1dpkJMSOEMeZs4GyA7u36A7A7c9i9Zg6U+1bq3j3sE5WUQP+fd4Kn4LJfbmBAsGAbQQXZ/1gtooIX5UF6ZWWwZkdHfsbzjDOfMuaYgRwxYkTixykiIiKSQmkTkI0xo4HJ0PAT/UDW2mnANIBhQ8faExf8mR47l5GdDeeeCwOG5MOxx0b0nL12d9NlDegQ5IQLkHmnzw0WkJs4SK+01LWNvFL9E6bn/YTplwARzmQnIiIi0lKlTUAGSoEBwBJf9bgdkG2M2d1aOybYHQrbwXkzjq1rbwhaBW5KU2ckg4gryC1GlC0WJSUwfXr9AWAtokouIiIiEqN0Sn7TgCc93/8WF5jPDbq2T0ztDeECcmuoIIc5SK/FtI+IiIiIxEnaBGRr7VagLrkZYyqB7dbaNQl70tZaQfafUAXCT/MmIiIi0sqkbfKz1l6T8CdRBVkBWURERCRAVqoHkFKttYIczammRURERFoZBWRo3RXkcKeaFhEREWllFJDBBWRrG9/emirICsgiIiIiQGsPyPn50KaNC8JbtjS+vTVUkBWQRURERBpo3QEZmm6zUAVZREREpNVRQG4qIGd6BXnnTqipgayszHkTICIiIhIjBeTWXEHWAXoiIiIijSggt+IK8ifvuvaKHTkKyCIiIiJ+GVIajUFrrCBv2kRN23YM31YLwJrNBSwu1ymlRUREREAV5LqAPP3pdZR/YN10b/6Lv4KcKQE5NxcmTAAge9sW2uJOOf2WOYiyshSOS0RERCSNZEjya76llZ3oC0x65VJ45dLgK2VKi4UxMGMGbNnCRx/BEUdAdTXsyG/H9NJUD05EREQkPbT6CvLbOZPZSIfQKxx2GGRnJ29AiWYMtGvH3pPa8Z+32nHZDe2YPl3tFSIiIiJ+rb6CPOiU/ej1yEaqqyEvj1YVFktKWs9rFREREYlUqw/IJSUuFJeVQWmpAqOIiIhIa9fqAzKokioiIiIi9Vp9D7KIiIiIiJcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiEfaBGRjTL4x5gFjzGJjTIUx5gtjzGGpHpeIiIiItC5pE5CBHGApMBHoAPwB+LcxZkAqByUiIiIirUtOqgfgZ63dAlzjWfQ/Y8xCYC9gUSrGJCIiIiKtTzpVkBswxnQHBgNfByw/2xgz0xgzc82aNakZnIiIiIhkrLQMyMaYXOBx4GFr7bfe26y106y1Y621Y4uLi1MzQBERERHJWGkXkI0xWcCjQDVwfoqHIyIiIiKtTNr0IAMYYwzwANAdONxauyPFQxIRERGRViatAjLwD2AYMNlauy3VgxERERGR1idtWiyMMf2BXwKjgVXGmErfZWpqRyYiIiIirUnaVJCttYsBk+pxiIiIiEjrljYVZBERERGRdKCALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4qGALCIiIiLioYAsIiIiIuKhgCwiIiIi4pFWAdkY09kY87wxZosxZrEx5hepHpOIiIiItC45qR5AgLuAaqA7MBp4yRgzy1r7dUpHJSIiIiKtRtpUkI0xhcAxwP9Zayutte8B/wFOSu3IRERERKQ1SacK8mBgp7V2nmfZLGCidyVjzNnA2b5vq4wxs5M0vkzVFVib6kG0cNqGsdH2i522Yey0DWOj7Rc7bcPYDInng6VTQG4HbA5Ytgko8i6w1k4DpgEYY2Zaa8cmZ3iZSdswdtqGsdH2i522Yey0DWOj7Rc7bcPYGGNmxvPx0qbFAqgE2gcsaw9UpGAsIiIiItJKpVNAngfkGGMGeZaNAnSAnoiIiIgkTdoEZGvtFuA54DpjTKExZj/gKODRJu42LSmDy2zahrHTNoyNtl/stA1jp20YG22/2Gkbxiau289Ya+P5eDExxnQG/gkcDKwDLrPWPpHaUYmIiIhIa5JWAVlEREREJNXSpsVCRERERCQdKCCLiIiIiHikXUA2xnQ2xjxvjNlijFlsjPlFiPWMMeZPxph1vsufjDHGc/toY8ynxpitvuvRSXsRKRbFNrzUGDPbGFNhjFlojLk04PZFxphtxphK3+X15LyC1Ipi+11jjNnh2T6VxphdPLdrHwy/DV8J2H7VxpivPLe31n3wfGPMTGNMlTHmoTDrXmyMWWWM2WyM+acxJt9z2wBjzNu+ffBbY8zkhA8+TUS6DY0xp/h+PzcbY5YZY/5sjMnx3F5mjNnu2QfnJuUFpFgU2+9UY0xNwO9xqed27YPht+E9AduvyhhT4bm9te6D+caYB3z/QyqMMV8YYw5rYv24/i1Mu4AM3AVUA92BqcA/jDHDg6x3NvBT3FRwI4GfAL8EMMbkAS8CjwGdgIeBF33LW4NIt6EBTsZto0OB840xxwes8xNrbTvf5UeJHHQaiXT7ATzl2T7trLULQPsgEW5Da+1h3u0HfAA8HbBaa9wHVwA34A5aDskYcwhwGTAJ6A/sAlzrWeVfwOdAF+BK4BljTHEiBpyGItqGQFvg17izmO2N25a/DVjnfM8+GNezdaWxSLcfQHnA38Eyz23aB8NsQ2vtOQF/B/9F47+DrXEfzAGW4s6o3AH4A/BvY8yAwBUT8rfQWps2F6AQ9091sGfZo8DNQdb9ADjb8/0ZwIe+r38ELMd3EKJv2RLg0FS/xnTahkHueztwh+f7RcDkVL+mdN1+wDXAYyEeR/tglPsgMACoAQZ4lrW6fTBgm9wAPNTE7U8AN3m+nwSs8n09GKgCijy3vwuck+rXlU7bMMj6lwD/9XxfBpyZ6teRrtsPOBV4L8Rt2gcj2IYB6xbiTpA20bOsVe+DAdvnS+CYIMvj/rcw3SrIg4Gd1tp5nmWzgGDVu+G+24KtNxz40vq2gs+XIR4n00SzDesYYwwwgcYnZnncGLPGGPO6MWZUfIealqLdfj8xxqw3xnxtjDnXs1z7YJT7IO7TjHettYsClre2fTAawf4OdjfGdPHdtsBaWxFwe2vYB2NxAI3/Dv7RGLPWGPO+t31A6uzp2z7zjDH/52lR0T4YvWOANcA7Actb/T5ojOmO+/8S7ARycf9bmG4BuR2wOWDZJqAoxLqbAtZr5wt6gbc19TiZJppt6HUNbn940LNsKq6q1x94G3jNGNMxHoNMY9Fsv38Dw4Bi4CzgKmPMCZ7H0T5YL5LXfjLwUMCy1rgPRiPY30Fw27o174PNYow5HRgL3OpZ/Hvcx7W9cSci+K8xZtcUDC9dvQPsAXTDhbsTAP/xLNoHo3cK8EhAcaXV74PGmFzgceBha+23QVaJ+9/CdAvIlUD7gGXtcR83hFu3PVDp26mieZxME/VrN8acjwsnP7bWVvmXW2vft9Zus9Zutdb+EdiIqzJnsoi3n7X2G2vtCmttjbX2A+A24OfRPk4Gas4+uD/QA3jGu7yV7oPRCPZ3ENy2bs37YNSMMT8F/ggcZq1d619urf3IWlthra2y1j4MvA8cnqJhph1r7QJr7UJrba219ivgOvR3sFmMMf2AUuAR7/LWvg8aY7JwbXrVwPkhVov738J0C8jzgBxjzCDPslEEL6d/7bst2HpfAyN91WS/kSEeJ9NEsw39FZPLgEnW2mVhHtviDuzLZFFtvwDe7aN9MLpteArwnLW2Msxjt4Z9MBrB/g7+YK1d57ttF2NMUcDtrWEfjIox5lDgPtwBoV+FWV37YNMC/w5qH4zcScD71newdxNazT7o+x/6AO6A72OstTtCrBr/v4WpbrgO0mj9JO5ow0JgP1wZfHiQ9c4B5uA+cujle6Hn+G7LAxYDFwH5uHcci4G8VL++NNuGU4FVwLAgt/Xz3TcPKMB9ZLYG6JLq15dG2+8o3AwVBhiPOyjvFO2DkW9D37ptfLcfpH2w7rXn+F7zH3GVkwIgJ8h6h/p+h3cHOgJv4TkYEvgQ1y5QAPwMV4EvTvXrS7NteBCwDjggyG0dgUP89/X9zdyC5wDUTL1Esf0OA7r7vh4KzAau9tyufTDMNvSsPxc4PWBZq90Hfa//Ht8+1C7MenH/W5jyFx/kRXYGXvDtAEuAX/iWT8C1UPjXM8CfgfW+y59pOGPAnsCnwDbgM2DPVL+2NNyGC4EduI8f/Jd7fLcNxx1UtsX3z2M6MDbVry3Ntt+/fNumEvgWuDDgcbQPhtmGvmUn4N48mIDlrXkfvAZXJfJersG9aagE+nnWvQT4Adf3/SCQ77ltAO4I+G24f76tZkaQSLchrrd9Z8DfwVd8txUDn+A+it2I+yd7cKpfW5ptv1t9+98WYAGuxSJX+2DUv8clvm1YFPAYrXkf7O/bZtsDfj+nJuNvofHdUURERERESL8eZBERERGRlFJAFhERERHxUEAWEREREfFQQBYRERER8VBAFhERERHxUEAWEREREfFQQBYRERER8VBAFhERERHxUEAWEREREfFQQBYRaWGMMcXGmJXGmKs9y0YaY7YbY6akcmwiIplAp5oWEWmBjDGHAP8FJgJfADOBj621p6VyXCIimUABWUSkhTLG/B04EpgBTABGW2srUzooEZEMoIAsItJCGWPygVnAIGBfa+1HKR6SiEhGUA+yiEjLNQDoC1hgl9QORUQkc6iCLCLSAhljcoEPgXnAR8DVwChr7ZKUDkxEJAMoIIuItEDGmJuBXwAjgU3AK0ABcJC1tjaVYxMRaenUYiEi0sIYYyYCvwFOttZutK7ScSqwO/D7VI5NRCQTqIIsIiIiIuKhCrKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIhwKyiIiIiIiHArKIiIiIiIcCsoiIiIiIx/8De92EZYtDKH4AAAAASUVORK5CYII= ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plot_predictions(reg, X_train, y_train, [0.0, 2.0, 0.0, 14.0], label='Single Decision Tree') ", "fig_id = 'dt_2DLinearReg' ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('DecisionTreeRegressor') ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show() ", "### ENTER CODE HERE ###" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Perform Grid Searches ", " ", "You will perform a series of grid searches, which will yield the optimal hyperparamter values for each of the three model types. You can compare the values computed by the grid search with the values you manually found earlier. How do these compare? ", " ", "You must perform a course-grained grid search, with a very broad range of values first. Then, you perform a second grid search using a tighter range of values centered on those identified in the first grid search. You may have to use another round of grid searching too (it took me at least three rounds of grid searches per model to ascertain the optimal hyperparameter values below). ", " ", "Note the following: ", " ", "1. Be sure to clearly report the optimal hyperparameters in the designated location after you calculate them! ", " ", "2. You must use `random_state=42` everywhere that it is needed in this notebook. ", " ", "3. You must use grid search to compute the following hyperparameters. Use the following hyperparameter values as the bounds of the ranges to be probed in your initial (course-grained) stint of grid searching. Note that these are NOT upper bounds for the further refined grid searches. If your initial search returns the upper bound of your values, you can (and should) search further above and below that value. You should examine these hyperparameters and their associated ranges, learn from this example, and use these values as a basis for your own future work when performing grid searches. This is how you will build your machine learning intuition. Keep in mind, these values also depend somewhat on this specific dataset! (For example, it is important to notice that I decided to probe up to 1,000 for n_estimators. 10 would have been far too small. 1,000,000 would have been far too large.) ", " ", " GradientBoostingRegressor: ", " ", " - `max_depth = up to 32` ", " - `n_estimators = up to 1000` ", " - `learning_rate = from .01 to 1` ", " ", " RandomForestRegressor: ", " ", " - `max_depth = up to 32` ", " - `n_estimators = up to 1000` ", " - `min_samples_split = up to 20` ", " ", " DecisionTreeRegressor: ", " ", " - `splitter = ["best", "random"]` ", " - `max_depth = up to 32` ", " - `min_samples_split = up to 20` ", " ", " ", "4. `learning rate` should be rounded to two decimals. ", "5. The number of cross-folds. Specify `cv=3` " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Perform Individual Model Grid Searches ", " ", "In this section you will perform a series of grid searches to compute the optimal hyperparameter values for each of the three model types." ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'learning_rate': 0.01, 'max_depth': 8, 'n_estimators': 250} " ] } ], "source": [ "# ----- ", "# Coarse-Grained GradientBoostingRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "from sklearn.model_selection import GridSearchCV ", "parameters = {'max_depth': [8, 16, 24, 32], 'n_estimators': [250, 500, 750, 1000], ", " 'learning_rate': [.01, .1, 1.0]} ", "reg = GradientBoostingRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'learning_rate': 0.01, 'max_depth': 4, 'n_estimators': 250} " ] } ], "source": [ "# ----- ", "# Refined GradientBoostingRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'max_depth': [4, 6, 8, 10, 12], 'n_estimators': [100, 150, 200, 250, 300, 350], ", " 'learning_rate': [.001, .005, .01, .02, .05]} ", "reg = GradientBoostingRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'learning_rate': 0.0175, 'max_depth': 2, 'n_estimators': 240} " ] } ], "source": [ "# ----- ", "# Final GradientBoostingRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'max_depth': [2, 3, 4, 5], 'n_estimators': [210, 220, 230, 240], ", " 'learning_rate': [.01, .0125, .015, .0175]} ", "reg = GradientBoostingRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "On this dataset, the optimal model parameters for the `GradientBoostingRegressor` class are: ", " ", "- `learning_rate = 0.02` ", "- `max_depth = 2` ", "- `n_estimators = 240`" ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'max_depth': 32, 'min_samples_split': 10, 'n_estimators': 750} " ] } ], "source": [ "# ----- ", "# Coarse-Grained RandomForestRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'max_depth': [8, 16, 24, 32], 'n_estimators': [250, 500, 750, 1000], ", " 'min_samples_split': [5, 10, 15, 20]} ", "reg = RandomForestRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'max_depth': 26, 'min_samples_split': 10, 'n_estimators': 250} " ] } ], "source": [ "# ----- ", "# Refined RandomForestRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'max_depth': [22, 24, 26], 'n_estimators': [200, 225, 250], ", " 'min_samples_split': [8, 10, 12]} ", "reg = RandomForestRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'max_depth': 25, 'min_samples_split': 8, 'n_estimators': 260} " ] } ], "source": [ "# ----- ", "# Final RandomForestRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'max_depth': [23, 24, 25], 'n_estimators': [250, 260, 270], ", " 'min_samples_split': [8, 10, 12]} ", "reg = RandomForestRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "On this dataset, the optimal model parameters for the `RandomForestRegressor` class are: ", " ", "- `max_depth = 25` ", "- `n_estimators = 260` ", "- `min_samples_split = 8`" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'max_depth': 16, 'min_samples_split': 10, 'splitter': 'random'} " ] } ], "source": [ "# ----- ", "# Coarse-Grained DecisionTreeRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'splitter': ['best', 'random'], 'max_depth': [8, 16, 24, 32], ", " 'min_samples_split': [5, 10, 15, 20]} ", "reg = DecisionTreeRegressor() ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{'max_depth': 6, 'min_samples_split': 10} " ] } ], "source": [ "# ----- ", "# Refined DecisionTreeRegressor GridSearch ", "# ----- ", " ", "### ENTER CODE HERE ### ", "parameters = {'max_depth': [2, 4, 6, 8, 10, 12], 'min_samples_split': [6, 8, 10, 12]} ", "reg = DecisionTreeRegressor(splitter='best') ", "clf = GridSearchCV(reg, parameters, cv=3) ", "clf.fit(X_train, y_train) ", "print(clf.best_params_)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "On this dataset, the optimal model parameters for the `DecisionTreeRegressor` class are: ", " ", "- `splitter = "best"` ", "- `max_depth = 6` ", "- `min_samples_split = 10`" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Visualize Optimal Model Predictions ", " ", "In the previous section you performed a series of grid searches designed to identify the optimal hyperparameter values for all three models. Now, use the `best_params_` attribute of the grid search objects from above to create the three optimal models below. For each model, visualize the models predictions on the training set - this is what we mean by the "prediction curve" of the model. Remember when plotting, to plot the training data not the testing data. ", " ", "### Create Optimal GradientBoostingRegressor Model" ] }, { "cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Score on test set is 0.6876 " ] } ], "source": [ "### ENTER CODE HERE ### ", "gt_reg = GradientBoostingRegressor(learning_rate=0.02, max_depth=2, n_estimators=240) ", "gt_reg.fit(X_train, y_train) ", "print('Score on test set is {:.4f}'.format(gt_reg.score(X_test, y_test)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plot Model Predictions for Training Set" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure best_gbrt_2DLinearReg " ] }, { "data": { "image/png": "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 ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plot_predictions(gt_reg, X_train, y_train, [0.0, 2.0, 2.0, 12.0], label='Ensemble Predictions') ", "fig_id = 'best_gbrt_2DLinearReg' ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('GradientBoostingRegressor: learning_rate=0.02, n_estimators=240, max_depth=2') ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Create Optimal RandomForestRegressor Model" ] }, { "cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Score on test set is 0.7136 " ] } ], "source": [ "### ENTER CODE HERE ### ", "rf_reg = RandomForestRegressor(max_depth=25, n_estimators=260, min_samples_split=8) ", "rf_reg.fit(X_train, y_train) ", "print('Score on test set is {:.4f}'.format(rf_reg.score(X_test, y_test)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plot Model Predictions for Training Set" ] }, { "cell_type": "code", "execution_count": 22, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure best_rf_2DLinearReg " ] }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAAFgCAYAAACmDI9oAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAAsTAAALEwEAmpwYAABesUlEQVR4nO3dd5wU9f3H8dfn7ugcXYpUFcGOKIpnULFgI4rGWGLBXmKLRo0NDFY0aqLGbhIbdn/2qFjRoIeKBXsBQZFepMMBd9/fH99Zbm5v7273buvt+/l47GN2Z2ZnvjM7u/uZ73zm+zXnHCIiIiIi4hVkugAiIiIiItlEAbKIiIiISIgCZBERERGREAXIIiIiIiIhCpBFREREREIUIIuIiIiIhChAlqQxszFmNi7T5ZDsZ2YrzGzTTJejocxsgpmdkulyhJnZA2Z2TfB8NzP7LtNlyjQzu9vMRme6HJJ64eM/BcvWf1weUYDcyJnZDDNbHQQkc4Mfj9aZLlcizKyPmblgGyKPKWkugzOzvqHXQ82sIijLcjP7zsxOTGeZcplzrrVz7sdMl6Oxc879zznXP/I6+D3Ypz7LMrO9zOwTM1tmZj+a2WnJK2nymNkJZjYxPM45d4Zz7uoUra/Kb0MmmNlFZvZl8Fs03cwuijHPn4JpK83sGzPrF5p2tJn9FEx7zsw6pHcLsk/wG/9LCpbbzMz+Hezv5Wb2mZkdEJoe6/9OJ3cZoAA5PxzknGsNbA8MBC7NbHHqrV0QWLV2zg1I9M1mVpTk8swO9msb4HzgPjPrX8d7EpaCcufU+iXzzKwJ8CxwD9AWOBL4u5kl/D2USkn8bhkwEmgP7A+cbWZHhdZzCnAyMBxoDfwWWBhM2xr/uR4HdAFWAXcmqVxSXREwE9gD/10aBTxpZn2i5gv/36Xk5E5qpwA5jzjn5gLj8YEyAGZ2iZlNC85kvzazQ0PTTjCziWZ2k5n9GtQ+hM90NzGzd4L3vg50Cq/PzA42s6/MbElwKXrL0LQZQa3H50Gtxb/NrIuZvRIs7w0za1/XNpnZxmb2gpktNrOpZnZqaNoYM3vazMaZ2TLgBDNrG6xrjpnNMrNrzKwwmL9vsD1LzWyhmT0RjH83WOSU4Gz+yKj96pxzLwOLge2C9xSE9u0iM3syXCtjZiODGoRFZjY6XLOXxHKbmf3DzOabr/X7wsy2Caa1NbOHzGxBUI5RZlYQ+tzfC967CBgTx+cwISjT+8E+etHMOprZI8G6Pwr/AVio1s38VY07zOy/wWf/gZltVsf6atu24Wb2aTB+ppmNCb0vUjtzYjDtVzM7w8x2Co7FJWZ2e2j+yL64Pdi/35rZ3rWU6yTztXO/mtl4M+tdV3lrWdaB5r+Ty4PP/MJg/FAz+8XMLgs+7xlmdkwNy9hQC2ZmDwO9gBeDz+gvta0/Sgf8ieDDwfH+EfANsFU8bw6Oj6uDfbnczF4zs05xvG+X4JhaYmZTzGxoaNoJ5muyI7Wmx5j/jbkbKAm2cUkwbzjtJLL//hJ8HnPM7JBgf39v/rfkstB6djaz0qAMc4JjoWkwLeZvg5mdav73aLH536eNQ8tzZnaWmf0A/FCfYyOac+5vzrlPnHPrnXPfAc8DvwnWVwD8FTjfOfd18PlNc84tDt5+DPCic+5d59wKYDTwOzMrjuPzSea+3DU4nnsGrwcE36Mt6ijDQPNXNpab/+1rHjX9t+ZraZcEx9J2oWkzzOzS4Hv2q5ndb2bNzawV8AqwsVXW4kY+w6bmfzuXm/9/G1TXfgpzzq10zo1xzs1wzlU4514CpgM7JrIcSQPnnB6N+AHMAPYJnvcAvgBuDU0/HNgYf7J0JLAS6BZMOwFYB5wKFAJ/BGYDFkwvBf4ONAN2B5YD44Jp/YJlDQOaAH8BpgJNQ+WahK+x6A7MBz7B13A3B94C/hrM2wdwQFGM7XsXX9vRHB/4LwD2CqaNCcp/SLB9LaisBWsFdAY+BE4P5n8MuDyYtzkwJLQeB/QNvR4K/BI8LwAOBiqAgcG4PwXb1yPYP/cAjwXTtgJWAEOApsBNQTn3SWa5gf2Aj4F2+BqmLUOf7UP4P9HiYP9+D5wc+tzXA+fgaztaAEcDn9dynE0IPt/N8LUiXwfL3CdYxkPA/bH2J/AAsAjYOZj3EeDxOo7r2rZtKLBtsD+2A+YBh0QdS3cH+2pfYA3wXLBfI8fiHlH74nz8cXwksBToENruU4LnI4J9sGWwHaOA9+sqby3bOAfYLXjeHtghtH3rqfzu7YH/rvUP7c9roo/T6N+D0LgltTwuCc33KHAW/regJNhPPeP8HZoATMP/LrQIXl9fx3u6B8fFgcFnOSx4vRH+e7AstM3dgK1Dn9nEqGVF75P1wBXBZ3oq/nfjUfz3YWtgNbBJMP+OwC7BZ9oHf2JwXi2/DXvha2d3CD6ffwLvRs3/Ov6ko0VtxwZwSW2fTw37zYBPgTOC172Cdf4JX3M5HbgSKAimPw9cHLWMFcCOcXyuyd6X1+J/+1vg/6vOrmP9TYGfqPx+/h7/2xn5rAfij9PB+OP2ePx3oFno+/Al0DP4PN6jhu9O6Ld5Df6YLATGApNC01+q5fN6qYZt6BIsc4uo36hZwC/A/UCneL5neiT3kfEC6JHiD9j/AKzAB68OeBN/6aam+T8DRgTPTwCmhqa1DJbRNfjRXQ+0Ck1/lMoAeTTwZGhaQfCFHxoq1zGh6f8H3BV6fQ7wXPA88oMR/rG5MPhRKweKQ+8bCzwQPB9D1T+mLkAZ0CI07g/A28Hzh4B7gR4x9kusALkiKEtZUI7zQtO/AfYOve6G/+Euwv+ZPBa1X9dSNUBucLnxf9Tf4/+QCkLjC4P1bRUadzowIfS5/5zgcTYBuDz0+mbgldDrg4DPYu1PfPDyr9C0A4Fv61hfzG2rYd5bgH9EHUvdQ9MXAUdGHYvnhfbFhpPCYNyHwHGh7Y4EyK8QnGSEjvlVQO9Eyht6/8/B59ImavxQqn/3ngRGh/Zn3AFyAp/xQfiTjfXB49QEj49RoddnAq/W8Z6L8TXW4XHj8UFOK/x37zBC34vQZ1ZXgLwaKAxeFwfHxODQ/B8TnFTFKNd5wLOxjuXg9b+Bv4Vet8Z/9/uE5t+rPsdynPv6SmAKlUHgrsE6/4sPwvsE6zs1mP4mQTAdWsaG3+o61pXsfdkkmP8L4FVC37sa3r871b+f74c+67uAq6Pe8x2VJ8AzwtuO/+2ZFuu7E4wbA7wRer0VsLoBn1UT4A3gnqjjZRD+v6IL8DQwvqHHhR6JP5RikR8Occ4V47/wWxBKhTB/qT9y+WkJsA1VUyXmRp4451YFT1vja51/dc6tDM37U+j5xuHXzrkKfO1F99A880LPV8d4HX0zYSfnXLvgcVOwjsXOueVRZQivY2boeW/8D9Kc0Pbeg685BF/LbcCHwaWzk6jdbOdcO/yl59vwf3ThdT0bWs83+CC6S1DuDeUK9uuiqGU3uNzOubeA24E7gPlmdq+ZtcF/vk2o+nnVtt/ilejnGTY39HxVHfPWtm2Y2WAze9t8+shS4Ayi0n8SLOssF/xzBX7Cf4bRegO3hj6jxfjPpXtt5a3FYfg/7J/Mp9CUhKbF+u7FKlNSBJe5H8fnuTbF1wz+xcyGJ7CYhD5j/P48PLI/g306BF+7uhJfm38G/nvx37ouxUdZ5JwrD56vDoYxjwEz62dmL5m/yXkZcB3Vj6ew6N++Ffjvd8zvVz2PjZjM7Gz8ZzTcOVcWtX1/c84tcc7NwP9+HBiMX4H/DQtrg69UiUfS9qVzbh3+ZGYb4Oao710sGxP7+xnRG7gg6hjqSdXvysyo99b1PYo+jptbPXLJg9SXh/GVFWdHxjvnVjjnJjufLjMvmLZvPCkvklwKkPOIc+4d/I/PTQDm8yPvw38BOwbB3pf4P/W6zAHaB7laEb1Cz2fjf5wI1mX4H6ZZ9d+CamYDHaJ+OHpFrSP8wzkTXxMbDrTbOOe2Bp+j7Zw71Tm3Mb7m7k6L4+704I/oYmBbMzsktK4DQutp55xr7pybhd93PSLvN7MWQMfoxSaj3M6525xzO+JrOvoBF+Ev/64j9PnUsd+yUg3bBv5Kxgv4y/9t8ekU8RzTNekeHL8RvfDHXrSZ+LSX8Gfewjn3fh3lrWn7PnLOjcCfCD2HryWOiPXdi1WmaouNHmFV75aPfkTyR7cBvnfOjXc+b/I7fI3kAdHLS6KZ+Brk8P5s5Zy7HiAoyzD81Zlv8b9lMbexge4Klr+5c64NcBm1H0/Rv32t8N/vGr9fNR0b5vPMa/x8wssITowvwV+5Cre+8B0+CAuvM/z8K2BAaDmb4lNDvq9lG+ur1n1pZt3x+dL3AzebWbM6ljeH2N/PiJnAtVHHUEvn3GOheXpGvTfyPUr4ODJ/D01Nn9crofkMf6WhC3BYcGJQk0g5FK+lmXZ4/rkFGGb+7vNW+C/fAgDzzZTFdXOIc+4nYDJwpZk1NbMh+EuwEU8Cw81sb/N3wF+AD/LeT9aGOOdmBssbG9xYsR3+Tu2Y7VQ65+YAr+F/eNuYv5FuMzPbA8DMDjezSOD6K37fVASv5wE1ttvrnFuLTyu4Ihh1N3CtVd6ktZGZjQimPQ0cZP6mlKb4y3Y1/uHWt9zmbzwbHOz/lfg8t4qgtufJoHzFQRn/XNN+y0Y1bVswuRh/ZWGNme2Mz59uiM7AuWbWxMwOx+eIvhxjvruBS823ChC5EfLwOMoba/uamr/prG3w57ksxvyR795u+FYJnopjW6odx67yTvlYj+uC2T4FNjff1JuZv4nyt8DnQXkjNz/2iaMM8RqH/57sZ2aFwXd8qJn1MH9D74gg+CzD14KGv6s9gu9WMhTj9/8K87XUf4yaHr1PHwNONLPtgwDvOuCDoOa2mtqODefcdbV9PqFlHBOsZ5iLaj4xuEL1BL7Gvzj4rTgNny8LPuf/IPNtZrcCrgKeiVyZM3+D4wOJ7LBa1Lgvg6DxAXzgeDI++K2r9YZSfLpP5Pv5O/y9DBH3AWcE+9fMrJX5m3jDlSpnBcdUB/y9HE8E4+cBHc2sbbwb55w7oJbPK3wyeRf+d+Qg59zq8DKCsvYPfuc74q9OTnDOLY23HJIcCpDzjHNuAT5n9Qrn3Nf4oK4U/2OwLf4mhXgdjb/5YTH+rP+h0Hq+A47F36CyEB88HxQEksn0B3xO3Wz8jWx/dc69Ucv8kUvEX+ODyafxNVAAOwEfBDUzLwB/Cv3ZjAEeNH+Z7ogalv0foJeZHQTcGizjNTNbjr9hbzCAc+4rfI714/g/gRX4G0nKYi20AeVug/+D+BV/6XARcGPwnnPwf8Y/AhPxta7/qWnlQbD2VS3lS7fatu1M4Kpgv19B1ZrX+vgA2Bx/HF8L/N45F50Sg3PuWeAG4HHzl4+/pLKGtbby1uQ4YEawrDPwrQ1EzA2WNRsf4JzhnPs2jm0ZC4wKjuML45gfAOfcNOAk/J/1MuAdfK72v4JZeuK3K2lXiIIT4BH4WsYF+NrAi/D/WwX4k7rZ+N+fPagMtt7C14rONbOFSSjKhfjfuuX4z/CJqOljCP02BL8/o/H7Zw7+xtWjqFl9jo1o1+BrqT8K1VjeHZp+Nv53Zjb+937D9z34PToDfxzNxwexZ4be25PE/hdqU9u+PBd/Mjo6SJk4EX+isVtNCwv+T36HzztfjE+7eSY0fTL+xsHb8ft3ajBv2KP4Cogf8TeSXhO891v8yc6PwWeblBSmoELidPxN5XNDn1fk+70pPv96Of43pAz/PydpFmmNQEQyxHzHLUvwlx2nZ7g4EmJmJ+BvwhuS6bJEmG/qbJxzrkcds6aNmY0CFjjn7sl0WSR5glr4KcB2daQB5CQzm4H/ftdWqSJ5Sh0AiGRAUMv8Jj614ib8XdszMlkmkfpyzqWka1/JrKCGdss6ZxRphNKaYmFmZ5vZZDMrC+c0mW8M/nXzjYovMLOnzKxbLYsSyXUj8Jc7Z+Mv3x/ldDmniiAnss6bk3KZ+VZHYm1jzI4/GpsgdSfW9mdTOo9kkJn1qul3wMx61b0EkfpJa4pFkEBfgW8YvYVz7oRg/AH4ZmDG4xPubwc2ds7tn7bCiYiIiIiQoRxk811+9ogEyDGm7wC843zbvSIiIiIiaZOtOci74+9CrsbMTsM3UUOrVq123GKLRNqGFxEREZHG5uOPP17onNsoWcvLugDZfFu2V+BzNKtxzt2L71aXQYMGucmTJ6exdCIiIiKSbczsp7rnil9WtYNsvvevV/DtuP4v0+URERERkfyTNQFy0Hj2G8DVzrmHM10eEREREclPaU2xMLOiYJ2FQKGZNce3WtEF3/vR7c65u2tZhIiIiIhISqW7BnkUsBq4BN8N8epg3Cn47hXHNMa2TkVERETyXWkpjB3rh9kurTXIzrkx+H7rY7kyfSURERERkXQpLYW994a1a6FpU3jzTSgpyXSpapY1OcgiIiIi0jhNmOCD4/JyP5wwIdMlql3WNfMmIumzdOlSFi5cyNq1azNdFJG81rRpUzp16kTbtm0zXRSRlBg61NccR2qQhw7NdIlqpwBZJE+tWbOGefPm0aNHD1q0aIGZZbpIInnJOcfq1av55ZdfaNasGc2bN890kUSSrqTEp1VMmOCD42xOrwAFyCJ5a8GCBWy00Ua0bNky00URyWtmRsuWLenUqRMLFiygZ8+emS6SSEqUlGR/YByhHGSRPLVmzRpat26d6WKISKC4uJg1a9ZkuhgiggJkkby1fv16iop0EUkkWxQVFbF+/fpMF0NEUIAskteUdyySPfR9FMkeCpBFREREREIUIIuIiIiIhChAFpFG4YEHHsDMYj7atWuX6eLVW2S7pk6dWut8M2bMwMx44IEHkrLeMWPGVNuHO++8M4888khSll+bCRMmYGZMCPUkMHToUIYm2HDqc889x9///ve4li8iEqY7dESkUXnqqafo0aNHlXG6GbH+Jk6cSGFhIYsXL+a+++7j2GOPpaysjJNOOimt5bjzzjsTfs9zzz3HG2+8wZ///Ocq43fYYQdKS0vZaqutklU8EWlk9K8hIo3K9ttvT9++fTNdjEZj8ODBG04w9t13X7bccktuueWWGgPksrIymjVrlvRyJDOYbdOmDbvsskvSlicijY9SLEQkr0RSFiZNmsQxxxxDmzZt2HjjjTn33HOrtEG7fv16Ro8ezWabbUbz5s3p1KkTQ4YMYeLEiVWWd++99zJgwIAN85x88sksXry4yjxmxqhRo7j55pvp3bs3LVu2ZPjw4cyfP5/58+dzxBFH0LZtW3r27MkNN9wQs9yzZ8/mkEMOoXXr1nTs2JGzzjqL1atX17m977zzDnvvvTfFxcW0atWK/fbbjy+//LIee87XxA8cOHBDukckVeGZZ57h1FNPZaONNqJLly4J7ZsFCxZw9NFH06ZNG9q1a8fIkSNZsmRJtXXHSrFYsGABZ555Jj179qRZs2b07NmT4447jrKyMk444QQefPBBZs2atSFNpE+fPlXKHU6xcM7xj3/8g/79+9O0aVO6devG2WefzbJly6qsM/JZ3nbbbWyyySYUFxezxx578NVXX1WZb/z48ey66660bduW1q1b079/f6666qoE97iIZIpqkEWkqmxpasq5er2tvLy8WluyBQUFFBRUrQ847rjj+MMf/sAzzzxDaWkpY8aMoX379lx55ZUA3HDDDfzjH//g2muvZfvtt2fZsmVMnjy5SoB3ySWXcPPNN3Puuedy4403MmvWLEaNGsWXX37J+++/T2Fh4YZ5H374YbbZZhvuvPNO5s2bx3nnncfIkSNZvnw5BxxwAKeddhpPPfUUl1xyCdtuuy0HHnhglfIee+yxHHHEEZx55pl8+OGHXHXVVaxcubLWnOP//ve/jBgxguHDhzNu3LgN27Xbbrvx+eef16vHtunTp1fL6T7nnHM44IADePjhhzecZMS7b373u98xZcoUrrvuOjbffHOeeOIJzjnnnDrL8euvv7LrrruyePFiRo0axXbbbcf8+fN5/vnnWbt2LaNHj2bBggV89NFHvPDCCwC11mxffvnljB07lrPOOouDDjqIr7/+mtGjRzNlyhTeeeedKsfPuHHj6N+/P7feeitr167loosuYsSIEXz77bcUFRXx448/cvDBB/P73/+eK664gqZNm/LDDz/w448/Jrq7RSRTnHM5+9hxxx2diNTP119/HXuCD00z/0jQ/fff74CYj+HDh1eb74orrqjy/uHDh7vNN9+8yutDDz20xvVNnz7dFRQUuCuvvLLK+IkTJzrAPfvss6Fditt8883dunXrNow7//zzHeCuvvrqDePWrVvnNtpoI3fCCSdUK+/pp59eZT3XXHONKygocN99992G8gDu/vvv3zDPZptt5vbaa68q71u6dKnr2LGj+9Of/lTjtjnn3F//+lcHuDVr1rh169a5efPmuTFjxjhgw3vffvttB7hDDjmkXvvmtddec4B77LHHqsy3//77O8C9/fbbG8btsccebo899tjwevTo0a6goMB98sknNW7D8ccf77p3715tfKTckeUvWrTINW3a1B1//PFV5nv44Ycd4J5//vkN4wDXt29ft3bt2g3jnnrqKQe49957r8rrpUuX1li2mtT4vRSRWgGTXRJjTKVYiEhVmQ+N6117DPDss8/y0UcfVXnccsst1eYbPnx4ldfbbrstP//884bXO+20Ey+//DKXX345EydOZO3atVXmf/3116moqOCYY45h/fr1Gx6DBw+muLiYd999t8r8w4YNq3Kz4BZbbAHAfvvtt2FcUVERffv2ZebMmdXKe8QRR1R5fdRRR1FRUcGHH34Ycz/88MMPTJs2rVr5WrZsSUlJSbXy1aR58+Y0adKELl26cN1113Heeedx/fXXV5nn0EMPrfI63n1TWlpKYWEhhx12WLVtq8trr73GTjvtxMCBA+PajtpMmjSJtWvXcuyxx1YrR1FREe+8806V8cOGDaNJkyYbXm+77bYAG46f7bffniZNmnDUUUfx9NNPM3/+/AaXUUTSSykWItKobLPNNnHdpNehQ4cqr5s1a0ZZWdmG15dddhnNmzdn3LhxXHfddbRu3Zrf//733HjjjXTq1GlD0FPTuhYtWlTldfv27au8btq0aY3jw7nQEeHc3vDrWbNmxVx/pHwnn3wyJ598crXpvXr1ivm+aJMmTaKwsJD27dvTq1evKoFhRLdu3WKuu659M2fOHNq3b19tmdHbWtMyBgwYENc21CWSNhO9HUVFRXTs2LFa3nSsYwfY8Ln17duX8ePHc8MNN2zIid5555254YYb2GOPPZJSZhFJLQXIIiIxNGnShIsvvpiLL76YuXPn8tJLL/HnP/+ZVatW8cQTT9CxY0fA12RGB7nAhunJMm/ePLbeeusqrwG6d+8ec/7I+seOHcs+++xTbXokQK/LjjvuWGczedFdJMe7b7p168avv/7KunXrqgTJkW2rTadOnWo8OUhUJOCdO3dulX28fv16Fi1aVC0gjseee+7JnnvuSVlZGe+99x5XXHEFw4cPZ8aMGXTq1Ckp5RaR1FGALCJSh65du3LKKafw8ssvb2gBYtiwYRQUFPDzzz8zbNiwlJfhySefZK+99trw+vHHH6egoIDBgwfHnL9///706dOHr776iksuuSTl5QuLd9+UlJRQXl7O//3f/1VJq3j88cfrXMe+++7LNddcw5QpU2qsSW7WrFlcLX3ssssuNG3alMcff5y99957w/gnnniC9evXJ9xBSXQZ9tprL1asWMGIESOYPn26AmSRHKAAWUQalc8++4yFCxdWGz9o0KCEOgwZMWIEAwYMYIcddqB9+/Z8+umnvPrqq5x++ukAbLbZZlx88cWcffbZfPfdd+yxxx40b96cmTNn8vrrr3PKKaew5557Jm27Xn75ZS666CL23XdfPvzwQ6688kpGjhzJ5ptvHnN+M+OOO+5gxIgRrF27liOOOIJOnToxb9483n//fXr16lWtA41kiXffDBs2jCFDhnD66aezcOHCDa1YxNMM3fnnn8+jjz7KPvvsw6hRo9h2221ZuHAhzz//PHfffTfFxcVstdVWLF68mLvuuotBgwbRvHnzDfnCYR06dOCCCy5g7NixtGrVigMPPJBvvvmGUaNGMWTIkGr56nW5++67effddznwwAPp2bMnCxcuZOzYsWy88cZss802CS1LRDJDAbKINCqHH354zPELFixIqOZu991356mnnuKOO+5g1apV9OrVi7/85S9cfvnlG+a57rrr2HLLLbnjjju44447MDN69uzJ3nvvXWPgWl/jxo3j5ptv5q677qJp06aceuqp3HTTTbW+58ADD+Tdd9/l2muv5ZRTTmH16tV07dqVXXbZhSOPPDKp5YsW77555plnOPfcc7n00kspLCzk4IMP5vbbb+eQQw6pdfnt2rXjvffeY9SoUVx//fUsWrSILl26sNdee21IHznllFOYNGkSl112GUuWLKF3797MmDEj5vKuvfZaNtpoI+6++27uvPNOOnbsyMiRIxk7dmy1JgLrMmDAAF555RUuvfRS5s+fT4cOHRgyZAiPPPIILVq0SGhZIpIZ5hpwt3imDRo0yE2ePDnTxRDJSd988w1bbrllposhIiH6XorUj5l97JwblKzlqZk3EREREZEQBcgiIiIiIiEKkEVEREREQhQgi4iIiIiEKEAWyWO5fJOuSGOj76NI9lCALJKnmjRpElcnCiKSHqtXr47ZlbdIJpSWwtixfpiP1A6ySJ7q3Lkzs2bNonv37rRo0aJad8Eikh7OOVavXs2sWbPo0qVLposjWaK0FCZMgKFDoaQk/evee29YuxaaNoU330x/GTJNAbJInmrTpg0As2fPZt26dRkujUh+a9KkCV26dNnwvZT8lukAdcIEv+7ycj+cMEEBsojkkTZt2ugPWUQky2Q6QB061AfmkQB96ND0rTtbKEAWERERySKZDlBLSnytdaZSPLKBAmQRERGRLJINAWpJSW4ExpFcbShulczlKkAWERERyTK5EqBmUjhXGzbvl8xlq5k3EREREck54VxtIKlNMakGWURERERyx7Rp8NJLHP1LGSsMyg1ucCS1px0FyCIiIiKSNlXaeP7+Qbj8cliyJP4FrFwJQG/g2mDUDRR/n8wyKkAWERERkdQ66yy47z4qHOy4HnYEDAesT3xZLVrAoYdC9+6V4268cWWyigoKkEVEREQklZYvh/vug3XrKACahiatad6W5rffDEceGf/ymjWD6G7Zb7wxGSXdQAGyiIiIiKTOhAmwbh0MHsykv73LfvtVtvH86uuFlAwpzHQJq1GALCIikgWq5GWqeS9pTMaP98P992eX3Zvy6lvZf6wrQBYREcmwcHuuTZv6TiKyNXAQSVgkQN5vPyA32nhWO8giIiIZFm7Pde3aSM9gIo3ArFkwdSq0bQs77ZTp0sRNAbKIiEiGDR3qa44LC/1w6NBMl0gkSRYs8MPevaEodxIXcqekIiIijVRJiU+ryPa8TJGErVjhh61bZ7YcCVKALCIikgVyIS9TGr+k3yy6fLkfFhcnYWHpowBZRERERFJzs2iO1iCnNQfZzM42s8lmVmZmD0RN29vMvjWzVWb2tpn1TmfZRERERPJZSm4WVYAcl9nANcB/wiPNrBPwDDAa6ABMBp5Ic9lERETyQmkpjB3rhyIRKblZNEcD5LSmWDjnngEws0FAj9Ck3wFfOeeeCqaPARaa2RbOuW/TWUYREZHGTG0u575UdSqTkptFlYPcIFsDUyIvnHMrzWxaML5KgGxmpwGnAfTq1SudZRQREcl5sS6jK0DOHak+wUn6zaI5WoOcLe0gtwaWRo1bClQ73XDO3eucG+ScG7TRRhulpXAiIiKNhdpczm0516lMjgbI2VKDvAJoEzWuDbA8A2URERFptNTmcm6LnOBEapCz/gRHAXKDfAUcH3lhZq2AzYLxIiIikkRqczl35dwJjnKQ62ZmRcE6C4FCM2sOrAeeBW40s8OA/wJXAJ/rBj0RERGRqnLqBCdHa5DTnYM8ClgNXAIcGzwf5ZxbABwGXAv8CgwGjkpz2URERNJCzaxJ1luyBGbOhGnT4JtvYOpUqKhIfDk5GiCnu5m3McCYGqa9AWyRzvKIiIikm5pZk6zmHPzlL3Dzzf55WLt20L69P2D//nfo0qXu5UVSLBoYIKeqabuaZEsOsoiISF5obM2spTtwkRQpL4dDD4XJk2HOHN/MSdeu/iyuaVNYtsyPX7IEpk/3Z3affw6dO9e+3EgNcgNykDNxUpktzbyJiIjkhcbUzFokcBk92g+VMpLDpk2DF1/0QXCbNvDf/8Ivv8CPP8K338KsWf7x2Wewww4wbx48/njdy01CikUmmrZTgCwiIpJGkVYIrr4699Mrcq5NXqnZwoV+2K0bzJ4N++1XdboZbLwxDBgAF17oxz39dN3LTUKAnImTSqVYiIiIpFlOtUJQi5xrkzfLxUpXSVsKSyRA3mEHaNWq9nmHD4dmzWDiRPjpJ5+L3LQpFETVu1ZUwMqV/nldy6xFJpq2U4AsIiIi9ZJzbfJmsVh5tpDG3NtFi/ywU6e6523TBvbd16dk9Onjx221FXzyiQ+cI8LBcXTwnKB0n1QqQBYREZF6ayy14ZlWU7pK2m7ojNQgd+wY3/znngvvveeD4LIy+Ppr+OEH2GabynlytIk3UA6yiIiISMbFyrNNa+5tIjXIQGmrfRh74SJK314D++zjR/78c9WZcjhAVg2yiIiISIbVlK6SthSWBGqQo9NBpu/Tmy7g85HDktQGciYoQBYRERHJAhvSVR5+GC66B5Yto+SOOyi5dLdq8yb95r0EapCj00G+W11DgJyENpAzRQGyiIiISLZYtw7OOQeWLvWvDzkELr3UB5uFhdChA98v7sQfrhvKL+u6JO/mvUgNchwBcnTrJRvv0gveQCkWIiIiIpIC//ufD47794fNNoOXX4aLLqoySz/gYzrwG95j6totYt68l3ANcwIpFtHpIH3LesM1KMVCRERERFLgxRf9cMQIuPxyuOQS30lHp06wfj0sXsyKCR/R8duPmcgQ3ne70XOzvwGbb1hEvbpmTvAmvSqtl0zv7YfhAHnuXJ8qAgqQRUREJHvEW4uYts4opHbOVQbIBx3k2xu+885qs7VeuZJlv9mfTlMmcnDFc/B0EzjiyQ3TYzUZV+vnWlFRGSB36JB4ubt390H87Nk+ReS662DsWN/8G0C7dokvM8MUIIuIiDRC8dYi1qu2UVJj2jT/6NCh9g+hVSvafDzBf3h77gnPPAO//AI9egD16OFw6VIfJLdtC02aJF7upk19N9SzZvla4zFj/Pi99vJdV59xRuLLzDC1gywiItII1dTxRH3nkzT44AM/HDLE35BXm8JCP9/vfuc/vFtu2TApkiN89dVxnvAk2klILL2DNIuTT/bDq67yKx83zudT5xgFyCIiIo1QvJ1MpLUzCqndhx/64c47x/+e887zw5tvhiuv9Gka+KD40kvjvBqQYP5xTJEAGaBzZzj//PovKwsoxUJERKQRqqnjifrOJ2lQnwC5pMTnKZ99tk9tKC+HY4/1LWDUVQsdkYwa5LPPhnnz/I2EF1yQkzfmhZkLzjRy0aBBg9zkyZMzXQwRERGRhlm71t+UV1YGixdD+/aJvf+pp+Coo3wuMcBZZ8Htt9c8//z5PpCeNMnfWLdmDRx3HDz0UP23IYPM7GPn3KBkLU8pFiIiIiKZ9sUXPjju1y/x4Bjg8MPh0Udh++19ixL33utv3Itl0SL4zW/g9dd9W8Vr1kBBgb+pTgClWIiIiDR+q1bB6tVVx5n5QGzFCt8aQnl51enNm8Nuu0GRQoW0iKRXDB5c/2UceaR/HHUUPPEEHH88bLGFD367dYOjj4Y+feDuu2HqVNhuO3j+ed9qRlERtGyZlE1pDHTUi4iINGa33AIXX+wv4Ufr29c38bVgQez3XnQR/O1vKS2eBOqTfxwl0p71gQdeyoAnnoC33vKPiFtv9TXV993nX//tbz5glmqUgywiItJY3X03/PGP/nl0BxBlZbBypX8+YICvYYxwDl57zd/kNWUKbLVVesqbz7beGr7+2jf1Vo8gObo9688ueZx+5d/4linKy+GRR2DyZP9Zfv21D4ynTfO1y41AsnOQVYMsIiKSbs7BRx/59IaKispHYaFvlSBZLQA88ogf3nornHtu1Wnl5fD2236dQ4f6lIuwM86Ae+7xrRO8+Wb16ZI8y5bBN99QUdSEm8YPYLfyxFsTiW7P+v+aHMWlV4RmOOggn1Lx9df+9WmnNZrgOBUUIIuIiKRTRQX84Q/w5JOxp7drByNGwO67w4knNiwwjdykdeCB1acVFsI++9T83muvhaef9kH0k0/63FZJjY8/Buf4pHx7LruyGU3HJt6jYZ295222Gbzyik+56NLFH1tSIwXIIiIi6VBe7nsZ++gjX4vXujXstJOvxYs8FiyATz6BBx/0jx139OkP9VFR4bv+BejePfH3d+wIY8f6msbzz/fRWq9e9SuLxLZuHcye7QNX4AN2rtKjYSIBclztWe++u39InRQgi4iIpMNHH/mgF3wLEc8/H7tZrU8+gXPOgfffh6++qn+APH++D8A6dYIWLeq3jJNP9u3iTpzoA6vrroPNN6+cvmSJj8qGDoX996/fOvLVCy/AyJH+JsnAxwU7U+Dq36NhSYk6ekkWBcgiIiLp8Mknfrjllr6ar3Pn2PPtsIOPjt5/H77/vv7rmznTD3v0qP8yCgrgxRd98PvBB3DMMbHn+9e/fC9q8fbclu9efx1+9zt/VaFrV1Y1a8cHMzfm+YqDKCj0DY8o0M0sZWeLiIikWGkpfHb/p/7F6afXHBxH9Ovnhw0JkCP5xz171n8Z4HOi33oL7rgD9twTBg2qfOyyi59n0SK/kRKfceN8cPzHP8Ls2dx6+jcMszdZ7NrjnN+dklkKkEVEJO+Ulvr02nTEdJHmt9ZP9gHyl00G1v2mZATIyahBjmjZEs480wfKH31U+Sgt9fnJ4FMGJD4LF/rh/vuD2YYb7AoL659eIcmlAFlERPJKJGAdPdoPUx0kT5gAFWXr2JYvABg/b/u63xQOkOvbX0GyapDrcvDBfvjii6ldT2MSqSLu2BGovMHu6qsTb71CUkMBsoiI5JXo9mInTEjt+oYOhQFNvqYZa5lqfdl1/zZ1v6ljR9+xx/LlPre3PpJZg1yb3/zGp2F8+23lOqV2UQEy+KD40ksVHGcLBcgiIpJXIpezCwp8E8OhGCUlSkrgkQt9ekW7PQfGHwA1NM0iXTXITZpA//7+uQLk+MQIkCW7KEAWEZG8UlLiWwkoLPRNBZ93XurTLPqWfwdAp923jv9NkQD522+BeuRNp6sGGaBrVz+cMyf168p15eW+eTyA9u0zWhSpmQJkERHJO4sWVfbunI40C6ZO9cNwG8J12XFHP/zXvyh9ryKxvOmZMxvWSUiiunXzQwXIdfv1V59X3r49FKm13WylAFlERPJO2lsN+OEHP+zbN/73nHyyDzw/+gh3+SgOKHuWXconsrbM1R7QOwenngrr18Nhh9W/k5BEKECOn9IrcoJOXUREJO/E1S1vsjhXvxrkVq3gmmvg5JPZ9Z2x7BqMPqhoPEOH7lvz+yZOhPHj/U1+t99e72InJMsC5NLSNH229aEAOScoQBYRkbyUtm55582DlSt9wJpozukJJ/gA+913fZfPwK1/+pFNayt35Oa8YcMqc4NT7NslXdkC+PXbuWQ6qzbSjN/atf7qQNY1m6YAOScoxUJERCSVIrXHiaRXRBQU+FSLBx+ECy8EYNNOy2p/z9KlftgmjubkGqi01HcGd9Llvgb550lzMt6hXrqb8UuYAuScoBpkERGRVGpIgBwWCXiXL699vmVBAN22bcPWV4dITe2aNdDN+QC5i5vD/RMyW2MbyS+P1CBnXa90kV70FCBnNQXIIiIiqZSsALm42A+XxVmDnOIAOVJT6xzMpzMVGJ2Zz9Ah68lkeJHW/PL6UA1yTlCALCIikgjn4OWX4e67YcYM31ZceXllu3HRzyNt3iarBrmuADkyPcUpFuGa2oLCJqws6ETxmgWU9F0AdEvpuuuStvzy+lCAnBMUIIuISNZKWWsE06bB66/752Y1D2ONe+YZeOmlxNbXqhUMGdKwMkdqkOtKsUhTDXJ0TW3xGd3g8wW+JYtumQ2Qs5oC5JygAFlERLJSyloj+PhjH9GtWFH/ZbRtC5ddBvvt5zt7KCjwj8LC2M/btoWWLRtW7qBGePqUZcwtrWVfpPEmvSo1td26weef+1Y0ttuu5jeZ+X2TrxQg5wQFyCIikpVitUZQJSicMQNOPBG+/NKnPVRUVA7Dz6PHlZf79++xB2yxhR8Hfhh+XtO4du3gggvS04VzyBc/FbMtMHfqcvbcE95+u4YgOU036VUTqTUeMaL2+QoKYOedfbeABx6Y+nJl0ty5MGWK79VwzRp/IEc6jVGAnNUUIIuISFYK57gWFsLPP/ta5ZIS4KOPfHAVaREgUUccAQ8/7FeQI557sw3bAm1YRlmZbxY5ZoCcphSLag4+GJ5+Glavrn2+8nKYNAnOP7/xBsgzZsChh8Jnn8WebqY0lCynAFlERLLL2rXw+OOULF/Ol2fChx/AJ6VldLtnFlPuK2OTEY6ub4zzNaX77gv33efTFwoKfOARSWuIPI8el6OX+Nc08TnIxcSZg5yGFIsqDj207vxo8J9b9+7w/fcwezZsvHHqy5Zud97pg+NWrWDHHaFPH9/ld7Nm/rHjjtClS6ZLKbVQgCwijV5Wdzsr1d11F5x3HgCbBo+jItPKgWeC54cfDo88Ak2apLuEGTHiuDbwiK9BbtoURo6sYcZMpVjEq00b2G03eOUVnydyzDGZLlFylZfDo4/656+9BrvuWvv8kpWyKkA2sz7AnUAJUAY8DZznnFufyXKJSO7K+m5npbrnnvPDgw+GHj2YOxeeeb6QmRXdKStqxTnnwCa7dPE1lkVZ9TeWUjvv7WuQ29hyJrztKCmx6jM5l7kUi0TsuWdSA+SsOgmeMMHnHG+6aRYURuor235Z7gTm4xtQbAe8DpwJ3JbBMolIDqvzRi/JLkuWwMSJPgXiwQehXTu6AgNLYekEHwBtkq+fX1ERtGhBwerVlAxYBbSqPk9ZGaxb588GmzVLexHjtueefvjCCzBqlO+vunv3ei2q2knwGy7YP7Vo2bKy6b5ke+wxPzz22NStIw7xnjRk1clFFsm2AHkT4Hbn3Bpgrpm9Cmyd4TKJSA7L+m5nparXXoP1630LE+3abRidqo4fci44KC72N8EtX+7zW6PlQu0xwMCB0KkTLFgA114Lt94KW23lx22ySe054h07wuab+xMG51j66GL+tGYpFQ7arlnOVgc+BEtn1r7+rl39jZq33JLcILaiorKN7N//PnnLTVC8V850ha1m2RYg3wIcZWYTgPbAAcDo8AxmdhpwGkCvXr3SXDwRyTVZ3+2sVPXyy374298mfdHRwXBOBgdt2sD8+T7PuGvX6tMzdYNeogoLfUct//uf3/HPPw8fflivRe0fPABwwFJ87XlNQXZ5uW9+7bbb4KSTYMCAeq03po8/hnnzoFcv2Gab5C03QfFeOUvlFbacO/mMkm0B8rv44HcZUAg8CDwXnsE5dy9wL8CgQYNcmssnIjkoq7udzUIZ/WP79ls/3GWXpC42VjCck+k3dfWml+036IVtv71/nHOO/9yXLPGtWsyspfbXOZ/f+9NPle1St23LrDUd+Hmm0auX0f0Pu8P++/tWS2paxsiRMG6cD9KTGSD/979+OHx4RtMr4r1ylqorbDl58hklawJkMysAXsUHv7sCrYH/ADcAf8lg0URE8kbG/9jmzvXDJLcRGysYzsn0m0jNcCQQjpYrKRbRttiiQW/vHjziYgYHHFAZIF94YWIrmzMHrr7a1xTPnVu1Le7Zs/1w+PDElplk8V45S9UVtpw8+YySNQEy0AHohc9BLgPKzOx+4BoUIIuIpEWy/tjqVQvtnA8+IHb6QAPECoZzMv0mEiDXVYOc7SkWmbbPPn747ru+h7vmzeN739q1vvWUDz6oeZ5u3SpvQsygeK+cpeIKW06efEbJmgDZObfQzKYDfzSzm/A1yMcDn2e2ZCIi+SMZf2z1roVessS/qbg49g1oDVBTMJxz6TeRFIvGVoOcbp07+9SKKVN8hx015St37OgPmKIiGD8epk/343v2hJtugo028gFxOJ2ie3ffSkYey8mTzyhZEyAHfoe/Ue9ifHPwbwHnZ7JAIiL5JBl/bPWuhY6kVyS59jgi54LhWOqqQVaAHL8TTvDdXdd0sgHw668wdWrVcT16+C61d945pcXLdbn+fcuqANk59xkwNMPFEBHJaw39Y6t3LXQkvSLJ+ceNSl01yEqxiN955/lWLNbX0hfZ99/7dIqKChg0yH8xCgszegOepEdWBcgiIpL76l0LneIa5EZBNcjJVdeJxC67JL1FFckNCpBFRCTp6lULrQC5bnXVIM+b54cdO6anPCKNVA2NBIqIiKRZipp4a1TqqkH+5hs/7NcvPeWJobQUxo71Q5FcpRpkERHJDilq4q1R6dzZD2N1puFcZUcrW26ZvjKFZLwdbZEkUQ2yiIhkB6VY1G3rrf3wiy+qT/vlF1i50jc9lqEUi1gtmIjkIgXIIiKSHZRiUbfevX0e8vz5lfnGEZHa4wb2StcQkRZMCgtzt4MIEVCALCIi2SCFveg1Kmaw7bb+eXQtciT/OEPpFVDZgsnVVyu9QnKbcpBFRCT1pk/3veNFcmgBXnkFXnoJfvrJB8eLFvlqx06dMlfOXLDttvD++z5AjnSZDJUBcgZrkCH3O4gQAQXIIiKSSj/9BIceCp9+6q+777GHD5QXLfJBXljHjnDzzTV3+yteTTXIGb5BT6QxUYAsIklVWtqwboqzZR2SJJdc4oPjtm39DWRvvVU5rXVruOACGDgQmjf3HTI0wg4ukn68RgLkzz+vHLd4MXz0kX8euZFPROpNAbKIJE06mnhSM1JZ5vvv/c11K1f6DyXs11/hiSf8B/XFF75m+OOPfb6xGey8M3Tpkplyp0lKjtfttvP78tNP4e23Yc894Z//9J/BfvtBz55JKbtIPlOALCJJE6uJp2QHr+lYh8Tp+efhkEPqnu+00yqDto03TmmRsk1Kjtd27eDii+G66+C3v/Wtfvzyi5922WUNXHj2ysUrR7lYZvEUIItI0kSaeIrUlqWiiad0rEPi9Nhjfrjllj4AbtbM1wyHtWsHY8aku2RZI2XH61VX+Vr5F1+EadP8uGHDYPfdk7SC7JKLV45yscxSSQGyiCRNpImnVNaYpGMdicjbGqL162H8eP/8hRegb9/MlidLpex4LSz0NfgzZvjqaTPfRnIjlYtXjnKxzFJJAbKIJFU6mnjKlmakcqmGKOmB/Pvvw5Il0K+fguM6pOx4NYNNNknBgrNPLl45ysUySyUFyCIi9ZQrNUS1BvL//S88+6wPtgoKKocFBbBunW86bPny6gtdsMAPf/vbtG2H5K9su3IUj1wss1RSgCwiUk+5UkNUYyD/1ltw0EG+VYn6OvLIJJVSpHbZcuUoEblYZvEUIIuI1FPW1RCtWgVTp1a+Li+HceM476kXObKiHAdYBWx8O/AvfPNszsFJJ/k2iCsq/MM5PzTz6RMbbRR7fZ06Qa9eadgwEZH0UoCco/L2xiCRLJOuGqJPnprGvEfeYECrqWzctaL6DKtX+zaHFy+uNqkFsGnkhQNmhyYecADccw8U6e9ARCRCv4g5KJduDBKRBK1aBf/6F/zwg3+9bh1LPvmR7T96gwLiSIXYbDNo2bLydb9+8Oc/x+6Qo6AA+vSp3jSbiEieiztANrPn8BflXnbOxai+kHTJlRuDRKQWq1fD5Mk+IF6/3t8Q9/XXcMcdMHt2lVnbAWU05VkO5QsbwF77N2XvvWMsc6edYLfdFPDmCV1JFEmdRGqQVwJPAEvN7AHgfufcDykpldQqV24MEpEa3Hyz7zxjxYrY03fYAY47zrd1W1DAd0u7sv/Vv2Hmuq40bQq/HQ0oIMprupIoklpxB8jOuWPMrA1wDHAicImZTcTXKj/lnFudojJKlKy7MUhE4vfGG3Dhhf75gAHQuTM0aeJzgNu1g6OP9j2iFRRseEt/4NE9M/edV01ldikt9edXZWX+XkpdSRRJPnP1bN7HzLYGTgHOAMrwtcu3OOe+SV7xajdo0CA3efLkdK1ORKS6lSvhvfd8tLJ+feXjxx/h0Udh0SKfDxUZv3q1f33VVTB6dKZLX6vSUnjoIfjPf3yRVVOZeZGa40hwXFDge/jW55IaOjnMHWb2sXNuULKWV6+b9MxsY2AE8FtgPfB/QE/gczO71Dl3U7IKKCKS1U49FR57LLH3HHIIXHppSoqTLJFAbM2aymaSVVOZeZF7UCLB8T77wGGH+fGgzyaZlMaS3xK5Sa8JPig+CRgGfAr8DXjMObcimOdg4CFAAbKI5IcPP/TDvfeG4mKfKlFUBK1a+chl4ECfSxwZX1QELVpktsxxiARikeDYTPc8ZIPoe1AOOwzOO09BXCrohvj8lkgN8hzAgEeBS5xzn8eY513g12QULBfo0otIjnjkER/hHX10cpe7bh3MmOGX/dJL0Lx5cpefQeFArKgITjwRRo7Ub12mRd+DoiAudXRDfH5LJEA+H38z3pqaZnDOLQE2aWihcoEuvYjkiDlzfIsQ4G9+q6lXuPqYMcNHJr17N6rgGHQzcDaL7pxGQVxq6DuQ3xJpxeLhVBYk1+isXSRHvPhiZZ7A668ntxY50q1z377JW2YWSVcvgVJ/CuJSS9+B/KWe9OpJl15EcsQLL1Q+f/XV5AbIkd7uGmmAnGpKU0sOBXEiyacAuZ501i6SA1au9O0OR4wf75taKyio0s5wvUVqkDffvOHLyjNKUxORbJaEf4j8VVLiW2rSj7pIlnrtNSgrY/nWg1nWtgfMn+875SgshBNOaPjyG3mKRSrFSlMTEckWqkEWkcxZsMBXJa5ZA7Nm+d4PfvnFd7JRVydG7drBbbfVftNdkF5x43cjWFdewFWMopByCnC+E4977vG9LMRj9mx/w1/YN0G/SKpBTpjS1EQkmylAFpHMOeQQeP/9+r+/Rw+48cbY08rLfdNrwHMVB/OF25rruRgz+Iqt2HLdN/D557DTTnWvZ/p02GILH81FM4NNN63/NuQppamJSDZTgCwimeEcTJnin++1F2y1FbRsCR07+mC0SZOa3zt7NpxyCvzrXzBmjO+UI2LWLF8DPW0aLFzImu6bMXXRVliZX6VzMNkGsSXfwOTJ8QXIL7zgg+POnX1QHjZ8eKNr4i1ddHOZiGQrBcgikhm//upvoisu9jfSmSX2/n//26dn7LMPHHQQXHSRX87BB/sb8QLNjziYNw83HnoI/vMfX7H8acFOHLfuYR8gx2P8eD+88UbfW4aIiDRqCpBFJDN+/tkPe/ZMPDgG+POf4fDDYdIk/3j4YZ+/vH49DBjgc4tbtoSzzqJkM19TOXKkv6T/246D4HTgo4/qXs+aNZV3kO27b+LlFBGRnKMAWUQyY+ZMP+zVq37vP+ww+N//4Kef4JJL4Ntv/fijj4Zx42IG3Rsu6a/eHs4shK++qmxrLHJTYHExHHooLFrkbyKcPx9Wr4btt4euXetXVhERySkKkEUkMyI1yPUNkM1gyBD/OPhg+PBDn4s8eHDdNdItWvhIeeJEeOut6tNffbX6uOHD61dOERHJOQqQRRq5rO2trKEBclhxsa8JTsRzz/kUiyZN/CPScch33/lpvXr51inMfKrGMcc0vJwiIpITFCCLNGJZ3VtZJMWiZ8/MrL9jR9h//+rjhwyBk09Of3lERCRrqCc9kUYsq3srS2YNsoiISBIpQBZpxCK9lRUWZmFvZQqQRUQkSynFQqQRy9reytav9x16mEH37pkujYiISBUKkEUaubT1VlZaCg8+CBUVdc87e7afr2tX316xiIhIFlGALJIH0tKSxV/+4ptNi1dREZxzTooKIyIiUn8KkCXlsraZsTyRtpYsFi3ywyuvrLtDjSZNYL/9YOONU1AQERGRhsm6ANnMjgL+CvQC5gInOOf+l9lSSX1ldTNjeSJWSxbhzyBpJzDLlvnhiSdmruk2ERGRJMiqANnMhgE3AEcCHwLdMlsiaai6gjNJvUhLFpGTlHBLFkk9gVm+3A/btGlgiUVERDIrqwJk4ErgKufcpOD1rEwWRhqutuBM0qO2liySdgLjXGWA3Lp1g8ssIiKSSVkTIJtZITAIeMHMpgLNgeeAi5xzq0PznQacBtBL7admvaxtZizP1NSSRdJOYFau9EFyq1a+0eU8oNx6EZHGy5xzmS4DAGa2Mb7G+GPgIGAd8DwwwTl3eaz3DBo0yE2ePDl9hRRphKoEeoMrfNvEZoktZPZs355x164wZ04qipmQVAevyq0XEckuZvaxc25QspaXNTXIQKSW+J/OuTkAZvZ3YBQQM0AWkTj88gs89BCUlcWcXBI8uH0q7PEUnHEG3HZbYuvIovzjdASvyq0XEWncsiZAds79ama/AOEq7eyo3hbJRStWwNSpcPDBMHNm/O/75z/h8MNh8GCYN8/XDBfU0St9pAWL4uL6lzdBNdUSpyN4VW69iEjjljUBcuB+4BwzexWfYnE+8FJmiySpEM8lcOV4NsAvv8CWW/ogGWDnneHAA2t/T8uW8PPPcPvtsNdePqe4vBzGjoVLLqn9vWmuQa6tljgdwaty60VEGrdsC5CvBjoB3wNrgCeBazNaIkm6eC6BK8ezgcaP98Fx69Y+2H3wQWjXru73rVnje8P77LPKcY88UneAnOYa5NpqidMVvKatC28REUm7Oq6bppdzbp1z7kznXDvnXFfn3LnOuTWZLpckV6zgpj7zSC3+F/Stc+218Pzz8QXHAM2bwyef+BrhZct8qxRffulrpGuT5hrkSC1xYWHsWuKSErj0UgWwIiJSP9lWgyx1Wb0aXnjBDwG23RZ23DGzZUpQPJfAlePZQJEAeciQxN9rVtmW8d57++Pt1VfhlFNg3TqoqPDTmjatbO0izTXISnEQEZFUUoCca664Am66qfJ18+b+RqosaD0gXvEENwqAGmD2bPjxRx+sbrddw5a1//4+QL7xRvj3v+GDD3xuMsCgQTBpkq/GzUArFkpxEBGRVMmLADmrb/b64AMfCS5e7F83b+5bEBgwoPq8S5fCPff453/4g6/V+/VX+OknX5OcQ+IJbhQA1dPEiX5YUgJFDfyKDx/uA+Dvv/evzSqr9idP9m0e9+iRkVYsREREUqXRB8hZfbPXE0/AUUdVH3/ttZW9kRUXwyGHQO/ePhd0+XLYc0949FEYNgzeeMPnh+ZYgJwLsvrEqjZTpvjh4MENX1avXj5d47vvoEMHf8Nf69YwcKC/kW/uXB8gZ1E7yCIiIg3V6APkrG7Q/9VX/XD48Mok2+nT4eGHKwOOJUvggQeqvu+CC/ywe3c/nDUrxQXNP1l9YlWXefP8MHJ8NFSsqvyuXf1w7lw/VA2yiIg0Io0+QM7qm71KS/3wr3+FnXaqHP/Pf1beCDV1qs8BXbXKv+7du7I92x49/LCuFgYkYVl9YlWXBQv8cKONUreO6ABZNcgiItKINPoAucE3e911F/zlL7B+ffVpBQUwenTdbcTGsnixv2zdvHn1fOOCgsqey7bYwj9iiQTIqkFOuqw+sapLJEDu3Dkliy8tBTe9G7uCapBFRKRRavQBMjTwZq/nnqvsjSyWJ5+sX4A8aZIfDhrkI7D6CALkae/+wvzSHKrhzAE53YrG/Pl+mIIa5EjqyWlrurIrMPezuXQF1SCLiEijkhcBcoNEUhtee61qm7K//AL9+lXmeyYqkl7RgMhrysLuDABWff8Le++dY3my0ebN8/u6qMjfoBh+FBVBkybQrFllu7tpkKpWNFJ+818Ka5AjqSeznU+xWPLtHB8gqwZZREQaEQXIdYkEyO3bQ4sWleN79vTD+fN9u7CJBm4ffuiHu+xS76K9/UMPBgDdmZV7ebIRzvkuz264oe55zaBlS9+Swvnn+8/Dufo9ysvhnXd8SwwVFZXjI2WKfsQaX9O43r1hhx18YL9+PXz1lW8ODfh1ZRPu/PgEHqs4svLmvwGrKo+zsPbtK1sziVdZmQ9Wi4ri7z0vAZHUkwVlXaECNi6IykFWgCwiIo2AAuS6RAKXVq2qjm/e3F9OXrbMt0XcoUNiy/3mGz9sQPNsg4d3ouy6pnTgV9o1WcnQoa3qflMN0tqk2axZcNZZft+tXu3TTYqKfKsL5eVVH+vXV94pt3YtrFwJL77oH9nqww/hqadiTmoPPMyrXMBY1q5uxiaHrISF31TelBm2/fbw8ceV+ejxCN+gl4La9kjqyedPd4W/Q5uVUTnISrEQEZFGQAFyXSIBcsuW1ad16eIDg/nzEwuQV66EmTN92sAmm9S7aCW7Gms27g6zp/P6A7MYWNKvXstJSZNmq1f7jk3mzvU3I4Kv+d1oI7j5Znj++cp5Cwpg3Dg48sjal1le7pd3ww0+qDar+1FQUPO0zTeHffetTN2IPCD2/PGMd863Qzx1amW5N9sM+vYFM2a8+AVdbruM7QnaKp6PryXu2LHqti5d6mu3J02CXXeNf7+nMP84oqQESrb2ATJz5/rgPpKnH+miWkREJIcpQK7LypV+GCtA7twZfvjBByU1tTQRS6RXsr59G9zTWfPNesDs6Qz88B5Y3Bd+85uEuxdOWpNm333nOz+ZORMeecQHyWE77ADvveenAfznPz4dYZNN4jtRKCz0tcy33VaPwqVRLTuvzz778NG+x/D1f6ezww6w7cAi2HLL6sfXBRfA3/8OTz9dGSCvWuUD5uXLfVN/TZpUX0GKW7DYoLjYp7isWgU//+xPDFq2TDwlREREJAspQK5LXTXIkPiNet9+64eJBNU16dPH93T297/71506+fIkcFm+pibNEk67OOMM/4aIzp19HuzWW8P778Mnn8DvfudPKLbcEk44Ia033WWLnYZ3ZqfhdQSwv/99ZYD8xz/6HOaRI3234gD33gunnlr9fZEa5FQHyGbQrRv8+CNcf70ft/POqV2niIhImihArk1FRWUtaPgGvYhIEBIJSmpSVlbZVnFhYWX+cf/+DS/jqFH+Zq61a+Hxx2HhQpg2zacPxClWk2YJp12sW1fZdN1VV/lAeOutK6e/+CIcfDC88op/nafBcdwGD/a15TNn+tZSok2eHDtATkcnIRFdu/oA+Z57/OsLL0z9OkVERNJAAXJt1qzxw+bNY9fIRgLk2mqQ//1vuPzy2PMkowa5Xz+49Vb/fNYsH4h+8klCATJUb9Is4bSLzz/3+6tfP995SrSDDoJrr4V33/W13KefnlD58k5BAVOPv5qKsddT4Mp9Ax6HHUC344bBiBGVed3R0lWDDFW7st5mm8oeHnNUWm9UFRGRrKYAuTY1tWAREUmxqKkG+d134ZRT/PNu3XygPXNmZa98yahBDhs4sDJAruuGtzok3JNcpPa4tmbrLrvMP9Ik1wOep1qfyOiCEykvh8ICuHoHuHRAkGIRSdOJls4a5L/8xZ84Ogd//nNOXxFIyY2qIiKSsxQg16a2G/Sg7hrku+7yw/PP9y03mME111TWsCY7QN5hBz/85JMGLyrhnuTiCZDTqDEEPDFPUnr29Ok+8+b5li7atvUB6ief+Jv3IjXL6ahBHjTIp/U0Akm7UVVERBoFBci1qe0GPai9Bnn+fPi///M1bOefX1m7dtFFMHGizxtu3z655Y0EyJ9+Wr/OS6LU1JNczJrZLAuQG0PAE/skpcCnsUyZ4oPhjh3hxBP9jZph6QiQG5GEr5iIiEijpgC5NnUFyLXVIF9/vb9x7be/rex1D3ybu6++mnBR4koX6NHD5/cuXOib3urdO+H1xFOOajWzvWf7dn9btmxQxyfJ1FgCnpgnKf37+wD5m2/gn//0nYl07OjzgAE23RR22intZc1lCV8xERGRRk0Bcm3iDZDnzvWtCoBv+WLiRPjHP3yLFaNGNbgYcacLmPka3Jdegueegz/9qcHrjhazZrZLEPDvtVeD23Wur+gTiEYd8ERu7rzhBh8kd+0KX3+d/CsSeaamKyYiIpJ/FCDXpq6b9Nq18xHrypWxa+zGjvXNdTVQQukCJ53kA+Q77oBzzkmsm+I4xKyZ/XvQdNsBByR1XfGq6QSi0QY8kdz1SHOBkab+REREJCnyJkCuV4sGdd2kZwZXXAHPPlt1fI8evs3fE06oX2GjJJQucNBBPqXjhx/gjTd8V8pJVK1mdqf18PrrfuL++yd1XfFqDPnGCdlzT98qypw5/ka5SEspIiIikhR5ESDXu0WDulIswLdxfPnlSSlnTRJKFygqgtNO8y1ljBtHafG+yU0zmD6dkgVfULLJKvgJGPc/35pCv34+9zUDGku+cdy6davseAZyunk1ERGRbJQXAXK9axjjCZDTJKF0gcMOg9GjWffiKwx7qpw16wrjOzFYtQrGjIEVK6CoiLmzypmybivan3U0O+/XHm6/Hc4917eQEWYGF1xQzy1ruEadb1wTBcUiIiIpkxcBcr1rGLMoQE7IFlvAppvS5Mcf2b7gQ96rKKn7xGDFCl/z/NhjG0Z1DR4/vHQr34+5in7XXuCD42HDoEMHP1NxMZx5pu+kJIMabb6xiIiIpF1eBMj1rmHM1QDZzDcvd9ttHFzwEpOspPYTg1NO8V1iA7RuDWPG8Mbbhbz6cgUj3QNs576Av/7BTz/jjMoOUEREREQaobwIkKGeNYyRm/RqasUimwUB8jldn6TijKvYY6/C2NvvHDzzjH/eqhU88QQceCCtdoU734L7y07iNjuXg/t/R/HQQfC3v6V1M0RERETSLW8C5HrJ1Rpk8NXFffrQYsZULun/LJT8PvZ8CxfCr7/6VImlSzfktlbWurdj06EPUZyk9IV6tSYiIiIikkZ5HyDXGrDlcoDcpInv1vqss/zwnXd8O8X77ec7MIn47js/7N+/2o1fyc7rrXdrIiIiIiJplNxeJNJs5UrfF0dpaf3eHwnYRo/2w2rLyeUAGeDEE32TYDNm+BYohg+Hs8+uOk84QE6xWK2JiIiIiGSbnA6Qv/uuluA2DnUGbLkeILdo4Tfqvvsq22p+6im/wRFpDJAjrYkUFuZJe8UiIiKSk3I6xcK5Gto2Xr7cd6RQVgbr1vnulrfd1qcdhNTZ/Fsu36QX0a+ffzgHjz4K06fDxx/Dzjv76WkMkPOyvWIRERHJOTkdIJv52LdKcDt3LmyzDSxaVHXmbbbh4ePfYNzrXTjsMN/kb50BW67XIIeZ+fzju++G8eMzEiCD2isWERGR7JfTAXL//jByZFRwe/HFPjju3Bm6dPG1xrNmwZdfsstFQ3iSv3PZayXgOnLa6VZ7wNaYAmSA/ff3AfIDD/h9VF4O06b54HnzzTNdOhEREZGsYC662+AcMqiw0E1u3brqyGXLoFkz+Oor2GwzP27+fH7cbBibrvh8w2xzm/em64hdYIcdfDfJ4ZYdIrbaCr75Br78ErbeOoVbkibLl/sThzVrqo7fYgu/nSIiIiI5yMw+ds4NStbycroGmYoKHxBHu+KKyuAYoHNn3rr+Q+46+zaO4RE25Ue6rvkJnvjJd4yxySZw+OHVl9PYapCLi316xQcfQFGRfxQW+rscRURERATI9RrkgQPd5OimJ4qKaryp7t574f/+Dw47tILTBk/x3SvfcQccdBC88EL1N3TuDAsWwLx5/rmIiIiIZJ1k1yDndoA8aJCbPHly/Rcwbx507+5zcOfMgU6d/PjZs+Hpp+FPf/Kvly+H6FQOEREREckKSrFIpi5dYN994ZVXoGtX3yQG+KbhIlq29O0Ji4iIiEheyO8AGeD88+Htt/2Na5EONJo08WkXW20Fu+0W+wY+EREREWmUFCAPGwYrVlTtXa6wUEGxSB1KS9Xpi4iINE4KkEEBseS0TASqpaW+8ZNIL5RvvqkgWUREGg8FyCI5LFOB6oQJfp0xu3oXERHJcQWZLkBjUFoKY8f6oUg6xQpU02HoUB+QFxZGdfUuIiLSCOR1DXIyLk3rUrNkUiRQjRx/6QpUS0r8sa4cZBERaYzyNkBOVmCrS82SSZkMVEtKdKyLiEjjlJUBspltDnwBPO2cOzYV60hWYJupGjyRCAWqIiIiyZWVATJwB/BRKleQrMC2sV9qVlNeIiIikm+yLkA2s6OAJcD7QN+65q9vAJfMwLax1uApv1pERETyUVYFyGbWBrgK2As4pYZ5TgNOA+jSZbMGBXCNNbBNFuVXi4iISD7Ktmbergb+7Zz7paYZnHP3OucGOecGFRW1y0gTV/lCTXmJiIhIPsqaGmQz2x7YBxgY73uKi2HxYt0glyqNPb9aREREJJasCZCBoUAf4GczA2gNFJrZVs65HWK9oVUrBXCppjQUERERyTfZFCDfCzween0hPmD+Y21vUgAnIiIiIsmUNQGyc24VsCry2sxWAGuccwsyVyoRERERyTdZEyBHc86NyXQZRERERCT/ZFsrFiIiIiIiGaUAWUREREQkRAGyiIiIiEiIAmQRERERkRAFyCIiIiIiIQqQRURERERCFCCLiIiIiIQoQBYRERERCVGALCIiIiISogBZRERERCREAbKIiIiISIgCZBERERGREAXIIiIiIiIhCpDzWGkpjB3rhyIiIiLiFWW6AJIZpaWw996wdi00bQpvvgklJZkulYiIiEjmqQaZ/KxJnTDBB8fl5X44YUKmSyQiIiKSHfK+Bjlfa1KHDvXbG9nuoUMzXSIRERGR7JD3AXKsmtR8CJBLSvzJwIQJPjjOh20WERERiUfeB8j5XJNaUqLAWERERCRa3gfIqkkVERERkbC8D5BBNakiIiIiUkmtWIiIiIiIhChAFhEREREJUYAsIiIiIhKiAFlEREREJEQBsoiIiIhIiAJkEREREZEQBcgiIiIiIiEKkEVEREREQhQgi4iIiIiEKEAWEREREQlRgCwiIiIiEqIAWUREREQkRAGyiIiIiEiIAmQRERERkRAFyCIiIiIiIQqQRURERERCFCCLiIiIiIQoQBYRERERCVGALCIiIiISogBZRERERCREAbKIiIiISIgCZBERERGREAXIIiIiIiIhCpBFREREREIUIIuIiIiIhChAFhEREREJyZoA2cyamdm/zewnM1tuZp+Z2QGZLpeIiIiI5JesCZCBImAmsAfQFhgFPGlmfTJZKBERERHJL0WZLkCEc24lMCY06iUzmw7sCMzIRJlEREREJP9kTYAczcy6AP2Ar6LGnwacFrwsM7Mv0122RqYTsDDThchx2ocNo/3XcNqHDad92DDafw2nfdgw/ZO5MHPOJXN5SWFmTYBXgGnOudNrmW+yc25Q+krW+GgfNpz2YcNo/zWc9mHDaR82jPZfw2kfNkyy91825SADYGYFwMPAWuDsDBdHRERERPJMVqVYmJkB/wa6AAc659ZluEgiIiIikmeyKkAG7gK2BPZxzq2OY/57U1yefKB92HDahw2j/ddw2ocNp33YMNp/Dad92DBJ3X9Zk4NsZr3xrVWUAetDk053zj2SkUKJiIiISN7JmgBZRERERCQbZN1NeiIiIiIimaQAWUREREQkJOsCZDPrYGbPmtlKM/vJzI6uYT4zsxvMbFHwuCFoBSMyfXsz+9jMVgXD7dO2ERmWwD68yMy+NLPlZjbdzC6Kmj7DzFab2Yrg8Vp6tiCzEth/Y8xsXWj/rDCzTUPTdQzWvQ9fidp/a83si9D0fD0GzzazyWZWZmYP1DHv+WY218yWmdl/zKxZaFofM3s7OAa/NbN9Ul74LBHvPjSz44Pv5zIz+8XM/mZmRaHpE8xsTegY/C4tG5BhCey/E8ysPOp7PDQ0Xcdg3fvw7qj9V2Zmy0PT8/UYbGZm/w7+Q5ab2WdmdkAt8yf1tzDrAmTgDnwbyF2AY4C7zGzrGPOdBhwCDAC2Aw4CTgcws6bA88A4oD3wIPB8MD4fxLsPDRiJ30f7A2eb2VFR8xzknGsdPPZNZaGzSLz7D+CJ0P5p7Zz7EXQMEuc+dM4dEN5/wPvAU1Gz5eMxOBu4BvhPbTOZ2X7AJcDeQG9gU+DK0CyPAZ8CHYHLgafNbKNUFDgLxbUPgZbAefhezAbj9+WFUfOcHToGk9pbVxaLd/8BlEb9Dk4ITdMxWMc+dM6dEfU7+BjVfwfz8RgsAmYCewBtgVHAk2bWJ3rGlPwWOuey5gG0wv+p9guNexi4Psa87wOnhV6fDEwKnu8LzCK4CTEY9zOwf6a3MZv2YYz33gb8M/R6Br7JvYxvVzbuP2AMMK6G5egYTPAYBPoA5UCf0Li8Owaj9sk1wAO1TH8UuC70em9gbvC8H75VoOLQ9P8BZ2R6u7JpH8aY/8/Ai6HXE4BTMr0d2br/gBOAiTVM0zEYxz6MmrcVsBzYIzQur4/BqP3zOXBYjPFJ/y3MthrkfsB659z3oXFTgFi1d1sH02LNtzXwuQv2QuDzGpbT2CSyDzcwMwN2A76KmvSImS0ws9fMbEByi5qVEt1/B5nZYjP7ysz+GBqvYzDBYxB/NeN/zrkZUePz7RhMRKzfwS5m1jGY9qNzbnnU9Hw4Bhtid6r/Do41s4Vm9l44fUA2GBjsn+/NbHQoRUXHYOIOAxYA70aNz/tj0My64P9for+fkILfwmwLkFsDy6LGLQWKa5h3adR8rYNAL3pabctpbBLZh2Fj8MfD/aFxx+Br9XoDbwPjzaxdMgqZxRLZf0/iO7bZCDgVuMLM/hBajo7BSvFs+0jggahx+XgMJiLW7yD4fZ3Px2C9mNlJwCDgptDoi/GXa7vjOyJ40cw2y0DxstW7wDZAZ3xw9wcgcj+LjsHEHQ88FFW5kvfHoJk1AR4BHnTOfRtjlqT/FmZbgLwCaBM1rg3+ckNd87YBVgQHVSLLaWwS3nYzOxsfnAx3zpVFxjvn3nPOrXbOrXLOjQWW4GuZG7O4959z7mvn3GznXLlz7n3gVuD3iS6nEarPMTgE6Ao8HR6fp8dgImL9DoLf1/l8DCbMzA4BxgIHOOcWRsY75z5wzi13zpU55x4E3gMOzFAxs45z7kfn3HTnXIVz7gvgKvQ7WC9m1gsYCjwUHp/vx6CZFeDT9NYCZ9cwW9J/C7MtQP4eKDKzzUPjBhC7Ov2rYFqs+b4CtgtqkyO2q2E5jU0i+zBSY3IJsLdz7pc6lu3wN/Y1Zgntvyjh/aNjMLF9eDzwjHNuRR3LzodjMBGxfgfnOecWBdM2NbPiqOn5cAwmxMz2B+7D3xD6RR2z6xisXfTvoI7B+B0HvOeCm71rkTfHYPAf+m/8Dd+HOefW1TBr8n8LM51wHSPR+nH83YatgN/gq8G3jjHfGcA3+EsOGwcbekYwrSnwE/AnoBn+jOMnoGmmty/L9uExwFxgyxjTegXvbQo0x18yWwB0zPT2ZdH+G4FvocKAnfE35R2vYzD+fRjM2yKYvpeOwQ3bXhRs81h8zUlzoCjGfPsH3+GtgHbAW4RuhgQm4dMFmgOH4mvgN8r09mXZPtwLWATsHmNaO2C/yHuD38yVhG5AbayPBPbfAUCX4PkWwJfAX0PTdQzWsQ9D838HnBQ1Lm+PwWD77w6OodZ1zJf038KMb3yMjewAPBccAD8DRwfjd8OnUETmM+BvwOLg8TeqthgwEPgYWA18AgzM9LZl4T6cDqzDX36IPO4Opm2Nv6lsZfDn8SYwKNPblmX777Fg36wAvgXOjVqOjsE69mEw7g/4kweLGp/Px+AYfC1R+DEGf9KwAugVmvfPwDx83vf9QLPQtD74O+BX4/9886ZFkHj3IT63fX3U7+ArwbSNgI/wl2KX4P9kh2V627Js/90UHH8rgR/xKRZNdAwm/D0uCfZhcdQy8vkY7B3sszVR389j0vFbaMEbRURERESE7MtBFhERERHJKAXIIiIiIiIhCpBFREREREIUIIuIiIiIhChAFhEREREJUYAsIiIiIhKiAFlEREREJEQBsoiIiIhIiAJkEREREZEQBcgiIjnGzDYyszlm9tfQuO3MbI2ZHZ7JsomINAbqalpEJAeZ2X7Ai8AewGfAZOBD59yJmSyXiEhjoABZRCRHmdktwMHAO8BuwPbOuRUZLZSISCOgAFlEJEeZWTNgCrA5sKtz7oMMF0lEpFFQDrKISO7qA/QEHLBpZosiItJ4qAZZRCQHmVkTYBLwPfAB8FdggHPu54wWTESkEVCALCKSg8zseuBoYDtgKfAK0BzYyzlXkcmyiYjkOqVYiIjkGDPbA7gAGOmcW+J8TccJwFbAxZksm4hIY6AaZBERERGRENUgi4iIiIiEKEAWEREREQlRgCwiIiIiEqIAWUREREQkRAGyiIiIiEiIAmQRERERkRAFyCIiIiIiIQqQRURERERC/h/uVtPr6zqoVAAAAABJRU5ErkJggg== ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plot_predictions(rf_reg, X_train, y_train, [0.0, 2.0, 2.0, 12.0], label='Ensemble Predictions') ", "fig_id = 'best_rf_2DLinearReg' ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('RandomForestRegressor: min_samples_split=8, n_estimators=260, max_depth=25') ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Create Optimal DecisionTreeRegressor Model" ] }, { "cell_type": "code", "execution_count": 23, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Score on test set is 0.6518 " ] } ], "source": [ "### ENTER CODE HERE ### ", "dt_reg = DecisionTreeRegressor(splitter='best', max_depth=6, min_samples_split=10) ", "dt_reg.fit(X_train, y_train) ", "print('Score on test set is {:.4f}'.format(dt_reg.score(X_test, y_test)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plot Model Predictions for Training Set" ] }, { "cell_type": "code", "execution_count": 24, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Saving figure best_dt_2DLinearReg " ] }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAAFgCAYAAACmDI9oAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAAsTAAALEwEAmpwYAABKeUlEQVR4nO3deZgcVdn38e+dmUwme0ISA4SEKDuEPSzDAyYQ1kQEBQFli7KKiIIPsgUIa0DlAXkFERciiwIRBBcQEBkQGGQTkFWUJRAgZF8nk+28f5zqSU1P71tVd/8+19XXTFdVV506Xd199913nTLnHCIiIiIi4vWIugEiIiIiInGiAFlEREREJEQBsoiIiIhIiAJkEREREZEQBcgiIiIiIiEKkEVEREREQhQgS2TM7EEzOz6H5Zaa2ecq0SaJnpmNNjNnZo3B/ZyOkygF7d006naUQq283sys1cxOjLodYWY23cwuD/7fy8zeirpN1aacz2v4+RFRgCwZmdl7ZtZuZkvMbKGZPW1mp5pZ0ceOc+4g59yvc1iun3PunWK2ZWavBR/8S81sjZmtCN0/v5h1J21nupmtDNY738weMbMtS7X+ehQ+Tsxsspk9GZ5fyQ81M5sa3MabWWuZt9Vtv4LX4+hg3uRybLcUrzfJzjn3d+fcFon7wXO7byHrMrObzewtM1ub6rgwszPN7BMzW2xmvzKzXkU0vSakei8p8fq/Y2bvmtkyM3vDzDYv17akPBQgSy4Ods71BzYGrgLOAX4ZbZPy45zbJvjg7wf8HTg9cd85d2ViuUTWskg/CLYzAphFGfqqRO0sZvtWii9JcRB1X4qUwMvAacCLyTPM7ADgXGAC/j38c8AlFW1dnQky3CcAk4B+wBeAuZE2SvJWEx9wUhnOuUXOuT8ARwLHm9kYADPrZWY/MrOZZjbbzG4ys96Jx5nZIWb2UpC9+K+ZHRhM7/ypzMw2NbPHzWyRmc01s7tCj+/8+drMBprZrWY2x8zeN7MpiUAtkREI2rIg+PZ+UKZ9Cv2cf4KZzQT+Fkz/RvCtf4GZPWRmG4ces2WQGZ4fZG2OSNNf7cDdwA6hx25oZvcE7X/XzM4IzettZr8OtvmGmX3fzD4MzX/PzM4xs1eAZWbWaGa7B1n9hWb2spmNDy0/2czeMZ/9f9fMjs6hr/cws+eCec+Z2R6hea1mdoWZPQUsx3/QZurbZjO73czmBe17zsyGh9Y1zcyeDY6L+81svTTraTWzE81sK+AmoMV8hn6hmZ0MHA18P5j2xxz6eaqZ/S5o22Jgcqb9yMPEoL/nmtkPLfQFIt3xZN61ZvZp0A//MrMx6farEEH/XR4cJ0vN7I9mNsTM7gi2+ZyZjQ4tH369TTezG8zsz8Fx9A8z2yTL9lLuUzBvkpn9M5j+gZlNDT0u8Vr8ejBvgflfq3Yxs1eC5/snoeUnm9lTZvaT4Hh908wmZGhXXs9Bln2caGavB30yy8z+N5g+3sw+NLPzg+PgPQtedynWMd6C17eZ3QaMAv4YPEffz7T9ZM65G5xzjwIrUsw+Hvilc+4159wC4DJyPOZDfXxt0P/vmH+PmBw8R59aqPwpy/N7ZPBaHBDcP8h8VntYljbsFzy3i4Ln35LmZ3qvdmZ2hiW9Li3Fe0lolYPzOd5TtLcHcDFwpnPudef91zk3P5/1SAw453TTLe0NeA/YN8X0mcA3g/+vBf4ArAf0B/4ITAvm7QosAvbDfyEbAWwZzGsFTgz+/y1wQbBMM7BnaFsO2DT4/1bg/mA7o4F/AycE8yYDq4CTgAbgm8BHgCW1Pbzd0cH6bwX6Ar2BQ4D/AFsBjcAU4Olg+b7AB8DXg3k74jMDWwfzpwOXh5a9DXg5uN8DeAG4CGjCB5jvAAcE868CHgcGAxsBrwAfJj0XLwEjg3aOAOYBE4N17xfcHxZsezGwRfDYDYBtMvV18PwtAI4N9u2rwf0hoX6bCWwTzO8J/Ak4N82xcwr+WOgTPB87AwNC65oFjAnaeg9we9Jz0pji+ZoMPJm0nc4+z7Gfp+KPk0ODZXsDXwMWZriNyvI6ccBjQR+Owh+XiTZnOp4OCNo6CP/BvxWwQar9KuI13BpsfxNgIPB60L59g/bcCtyS5vU2HX9M7RosewdwZ5btZdqn8cC2Qb9vB8wGDk163m/CH5f74wO++4DP4I/3T4FxoWNhNXAm/lg8Ev9es16K46ag5yDDPn4M7BX8PxjYKbR/q4H/A3oB44BlrHsddj6nwbLJr+99k7azMMOt2+sOeBKYnDTtZeDI0P2hQT8PyeHYSfTx1/Gv4cvx7wE3BPu3P7AE6Jft+Q3m3xH0wRD8e/MXsmx/aLD+w4Pn+MygPVmf1xxel5NJ/V6S9njHvyenez5uDJYZFWz3O/jPinfxGfsexb6WdavsLfIG6BbvW6o37WD6M/ggy/AfAJuE5rUA7wb//wy4Ns26W0NvVrcCNwMbpVjOAZsGb9ArCYLRYN4pQGvw/2TgP6F5fYLHrp9hu6ODZT4Xmv8gQdAd3O+Bz5hujP8Q/nvS+n4GXBz8Px3/ob4QWBu8OW4XzNsNmJn02PMIghNCQVxw/0S6f4B+I3T/HOC2pPU9hM8Y9Q3acBjQO2mZlH2ND4yfTZrWRvCBG/TbpXkcO98Ank7sf4rn4KrQ/a2D57aB4gPkbP08FXiixK8TBxwYun8a8GgOx9M++A/t3Un6AE3eryLa1gpcELp/DfBg6P7BwEvJr7dQG34RmjcReDPL9tLuU4plryN4fwg97yNC8+fRNbi7B/hu6Fjo8gUYeBY4NsVxU9BzkKHdM/HvPQOSpo/HB3B9Q9PuBi5Mfk7JIUAu4LlOFSD/N+nY7Bn08+gc1jcZeDt0f9vgscOTnqMdsj2/wf1BQd/9C/hZDts/DngmdN+AD3N5XkPHcrrX5WRSv5fkdbynaPMewXb/HOzv6OD4OqmY51a3yt9UYiGFGgHMx2cr+wAvBD/BLQT+EkwHn+38bw7r+z7+ze9Z8yfUfSPFMkPxb+7vh6a9H7Ql4ZPEP8655cG//XLY/geh/zcGfhzan/lB20YE83ZLzAvmHw2sH3r8j5xzg/BvjO3AFqH1bpj02POB4cH8DZPaEf4/XTu/krS+PfHZr2X4YP5U4OPgJ8PEyYLp+npDuvYtdO/fVG1K5zZ8wH6nmX1kZj8ws55p1vU+/rkdmsf608nWz8nbLpXk/dkw1J6Ux5Nz7m/AT/AZuU/Nn2w1oAxtmx36vz3F/UyvkU9C/y/PsiyZ9snMdjOzx8yXvizCH5/Jz3k+bZ3lgqgkEO73sFI/B4fhg6f3zZcrtYTmLQhef9naVClLgfD+JP5fkuPjk/sf51zK5yTb8+ucWwjMwP9ydE0O2+7ynhg817m+Vyeke12mk9fxnkJ78PcHzrmFzrn38EmUiXmuRyKmAFnyZma74N+AnsSXF7Tjf74fFNwGOn+SGvg3p6w1XM65T5xzJznnNsRnZm607sNmzcX/NL5xaNoo/E/1xQp/yH4AnBLan0HOud7OuaeDeY8nzevnnPtmin2aif+Z7cfma7I/wGfWw4/t75xLvHF+jC+tSBiZQztvS1pfX+fcVcH2H3LO7Ycvr3gT+HkwPV1ff0TXvoXu/evIkXNulXPuEufc1visyhfwGaFU+zcK/9xmO5El1faTp2Xr526PMbOjbd2oJqluo7K0K9X+fBRqT7rjCefc9c65nfFZ9M2BszPsa1XIsE+/wZdjjXTODcSXU1jqteRkhJmFHx/u97BCn4N0+/ecc+4QfOnHffgsccJgM+ubQ5u6rTZ5QpZjMtfRd14Dtg/d3x6Y7Zybl+Pj85Hx+TWzHfC/LP0WuD6H9X1M6HUVPNfh11nG5zWQ7nWZ9+vLuo6GlHy7KVjsLfyvYeH1V+1ruZ4pQJacmdkAM/sCcCe+XvRfzrm1+MDrWjP7TLDcCPNnToMfweHrZjYhODlihKUY9szMvmJmieBwAf4NZW14GefcGvwH0RVm1j84GeMs4PYS7+pNwHlmtk3QtoFm9pVg3p+Azc3sWDPrGdx2CU766MY59wj+Dflk/M+/S8yfaNfbzBrMn5C1S7D43cF2B5vZCOD0LO28HTjYzA4I1tVs/sSfjcxsuPmTI/sCHfgs0tpgf9L19QPBvn3N/AmAR+IDhj/l132eme1tZtuaWQO+HnoVXZ/TY8xsazPrA1wK/C54jjOZDWxkZk1J08InDGbr526cc3e4daOapLrNzGGXzw6eu5H4L0aJkx/THk/BsbOb+cz6Mnx5TqKPkvcrJVt3ctvoHNpYdln2qT8w3zm3wsx2xdd+F+MzwBnB6/Ar+FrUB1IsV+hzkGr/moIvVAOdc6vwx3by8pcEy+2F/2I4I4d96fZ8Zzkmw6PvNJlZMz4Y7Rm8FyQ+328FTghea4PwdbrTQ49ttdDJdEVK+/wG7bsd/2vO1/Ffbk7Lsr4/A9uY2ZfNjzZzBl1/rcv0Xp2Q7nWZ6r0kIxcaDSnF7dRgmeXBNr4ffE5thH//L+h9VKKjAFly8UczW4L/tn4B/gSUr4fmn4M/UeIZ86MC/JWgrMA592yw7LX4E2gep3uWEmAX4B9mthSfgfiOSz0W67fxH2Lv4DPYvwF+VewOhjnnfg9cjS8NWAy8ChwUzFuCPzHlKHzg+0mwbKZxRX+IL2toxH9Y7oCvTZ4L/AJ/4hT4IPHDYN5fgd/hg9t07fwAf5LK+cAc/PNzNv513QP/5eEj/M+O4/AnLUKavg4ySl8AvoevK/w+/iSatFld8xfxSJfJWj/Yh8XAG/jn/rbQ/NvwH9Sf4E/KOoPs/obPiH1iZol2/RLY2vzPrPcFQXamfi6X+/Ene72E/2D/JWQ+nvA/d/8c/0XlfXy//zDVfmXY7sjgsaX4JaUUMu3TacClwfvJRXTNvBbiH8Bm+Of4CuDwVJnRIp6DdI4F3gvWdSq+zCrhk2BdH+FP8jrVOfdmDvsyDZgSPN//m8PyYQ/jf8nbA39+QTvweQDn3F+AH+BPVpuJ38eLQ48dCTyV5/bSyfT8TgM+cM791DnXARwDXG5mm6VbWfDe8xX8Cczz8M/1U6H5mZ7XhJSvS1K/l5TK6fikxEf48zhK/jkl5WfOKfMvEkdm9k3gKOfcuKjbUmrmL7Jxu3PuF1G3pdqZ2RRgjnPuZ1G3pZLMXxDjROfcnlG3JcH8MIu3O+c2yrJoLATZzbudc3tkXbgKmZkDNnPO/Sfqtkj10QD5IjFhZhvgf2Jtw2dKvoc/eUgkLeecLo0rBXHOfYjPOotIkoqWWJjZ6Wb2vJl1mNn00PTdbd2FF+aY2YwgWBCpJ034s52X4H/+ux+4MdIWiaRgZntZmpOVom5bqVj6E7JSXvhDClMPx5JUp4qWWJjZl/EnMxyAH5t1cjD9IPxQKg/hx5D8CbChc+7AijVORERERISIapDN7HL8RQomp5m/E34orf4VbZiIiIiI1L241iB/Hn92aTdmdjJ+yBT69u2785ZbdhsxTERERETqyAsvvDDXOTcs+5K5iV2AbGbb4YeHOSTVfOfczfhhbBg7dqx7/vnnK9g6EREREYkbM0u+EmxRYjUOsvmreT2IH5f171G3R0RERETqT2wCZPNXRfsrcJlz7rZsy4uIiIiIlENFSyyCS0U2Ag1AQ3DpydXAcPywVj9xzt2UYRUiIiIiImVV6QzyFPwlMM/FX2ayPZh2Iv4CCVM1BqKIiIhI7Wlrg2nT/N+4q2gG2Tk3FZiaZvYllWuJiIiIiFRKWxtMmAArV0JTEzz6KLS0RN2q9GJTgywiIiIitam11QfHa9b4v62tUbcos9gN8yYilbNo0SLmzp3LypUro26KSF1rampi6NChDBw4MOqmiJTF+PE+c5zIII8fH3WLMlOALFKnVqxYwezZs9loo43o3bs3ZhZ1k0TqknOO9vZ2PvzwQ3r16kVzc3PUTRIpuZYWX1bR2uqD4ziXV4ACZJG6NWfOHIYNG0afPn2ibopIXTMz+vTpw9ChQ5kzZw4jR46MukkiZdHSEv/AOEE1yCJ1asWKFfTr1y/qZohIoH///qxYsSLqZogICpBF6tbq1atpbNSPSCJx0djYyOrVq6NuhoigAFmkrqnuWCQ+9HoUiQ8FyCIiIiIiIQqQRURERERCFCCLSE2YPn06ZpbyNmjQoKibV7DEfv3nP//JuNx7772HmTF9+vSSbHfq1Knd+nDXXXfljjvuKMn6M2ltbcXMaA1dSWD8+PGMz3Pg1Pvuu4//+7//y2n9IiJhOkNHRGrKjBkz2GijjbpM08mIhXvyySdpaGhg/vz5/PznP+eYY46ho6ODb3zjGxVtx4033pj3Y+677z7++te/ctZZZ3WZvtNOO9HW1sbWW29dquaJSI3Rp4aI1JQddtiBTTfdNOpm1Izddtut8wvG/vvvz1ZbbcV1112XNkDu6OigV69eJW9HKYPZAQMGsPvuu5dsfSJSe1RiISJ1JVGy8Mwzz3D00UczYMAANtxwQ84444wuY9CuXr2aCy+8kE022YTm5maGDh3KnnvuyZNPPtllfTfffDPbb7995zInnHAC8+fP77KMmTFlyhSuueYaNt54Y/r06cOkSZP49NNP+fTTTzniiCMYOHAgI0eO5Oqrr07Z7o8++ohDDz2Ufv36MWTIEL71rW/R3t6edX8ff/xxJkyYQP/+/enbty8HHHAAr776agE95zPxO+64Y2e5R6JU4d577+Wkk05i2LBhDB8+PK++mTNnDl/72tcYMGAAgwYN4rjjjmPhwoXdtp2qxGLOnDmcdtppjBw5kl69ejFy5EiOPfZYOjo6mDx5Mr/+9a+ZNWtWZ5nI6NGju7Q7XGLhnOPaa69liy22oKmpiQ022IDTTz+dxYsXd9lm4rm8/vrr+exnP0v//v0ZN24cr732WpflHnroIfbYYw8GDhxIv3792GKLLbj00kvz7HERiYoyyCLSVVyGmnKuoIetWbOm21iyPXr0oEePrvmAY489lq9+9avce++9tLW1MXXqVAYPHswll1wCwNVXX821117LFVdcwQ477MDixYt5/vnnuwR45557Ltdccw1nnHEGP/zhD5k1axZTpkzh1Vdf5emnn6ahoaFz2dtuu40xY8Zw4403Mnv2bL773e9y3HHHsWTJEg466CBOPvlkZsyYwbnnnsu2227LxIkTu7T3mGOO4YgjjuC0007j2Wef5dJLL2XZsmUZa47//Oc/c8ghhzBp0iRuv/32zv3aa6+9eOWVVwq6Ytu7777brab729/+NgcddBC33XZb55eMXPvmy1/+Mi+//DJXXnklm222GXfddRff/va3s7ZjwYIF7LHHHsyfP58pU6aw3Xbb8emnn3L//fezcuVKLrzwQubMmcNzzz3HH/7wB4CMme0LLriAadOm8a1vfYuDDz6Y119/nQsvvJCXX36Zxx9/vMvxc/vtt7PFFlvw4x//mJUrV3L22WdzyCGH8Oabb9LY2Mg777zDF7/4RQ4//HAuuugimpqaePvtt3nnnXfy7W4RiYpzrmpvO++8sxORwrz++uupZ/jQNPpbnm655RYHpLxNmjSp23IXXXRRl8dPmjTJbbbZZl3uf+lLX0q7vXfffdf16NHDXXLJJV2mP/nkkw5wv//970Ndittss83cqlWrOqedeeaZDnCXXXZZ57RVq1a5YcOGucmTJ3dr7ymnnNJlO5dffrnr0aOHe+uttzrbA7hbbrmlc5lNNtnE7bPPPl0et2jRIjdkyBD3ne98J+2+OefcxRdf7AC3YsUKt2rVKjd79mw3depUB3Q+9rHHHnOAO/TQQwvqm4cfftgB7re//W2X5Q488EAHuMcee6xz2rhx49y4ceM671944YWuR48e7sUXX0y7D8cff7wbMWJEt+mJdifWP2/ePNfU1OSOP/74LsvddtttDnD3339/5zTAbbrppm7lypWd02bMmOEA99RTT3W5v2jRorRtSyft61JEMgKedyWMMVViISJdRR8aF5w9Bvj973/Pc8891+V23XXXdVtu0qRJXe5vu+22zJw5s/P+LrvswgMPPMAFF1zAk08+ycqVK7ss/8gjj7B27VqOPvpoVq9e3Xnbbbfd6N+/P0888USX5ffbb78uJwtuueWWABxwwAGd0xobG9l000354IMPurX3iCOO6HL/qKOOYu3atTz77LMp++Htt9/mv//9b7f29enTh5aWlm7tS6e5uZmePXsyfPhwrrzySr773e9y1VVXdVnmS1/6Upf7ufZNW1sbDQ0NHHbYYd32LZuHH36YXXbZhR133DGn/cjkmWeeYeXKlRxzzDHd2tHY2Mjjjz/eZfp+++1Hz549O+9vu+22AJ3Hzw477EDPnj056qij+N3vfsenn35adBtFpLJUYiEiNWXMmDE5naS33nrrdbnfq1cvOjo6Ou+ff/75NDc3c/vtt3PllVfSr18/Dj/8cH74wx8ydOjQzqAn3bbmzZvX5f7gwYO73G9qako7PVwLnRCu7Q3fnzVrVsrtJ9p3wgkncMIJJ3SbP2rUqJSPS/bMM8/Q0NDA4MGDGTVqVJfAMGGDDTZIue1sffPxxx8zePDgbutM3td069h+++1z2odsEmUzyfvR2NjIkCFDutVNpzp2gM7nbdNNN+Whhx7i6quv7qyJ3nXXXbn66qsZN25cSdosIuWlAFlEJIWePXtyzjnncM455/DJJ5/wpz/9ibPOOovly5dz1113MWTIEMBnMpODXKBzfqnMnj2bbbbZpst9gBEjRqRcPrH9adOmse+++3abnwjQs9l5552zDpOXfInkXPtmgw02YMGCBaxatapLkJzYt0yGDh2a9stBvhIB7yeffNKlj1evXs28efO6BcS52Hvvvdl7773p6Ojgqaee4qKLLmLSpEm89957DB06tCTtFpHyUYAsIpLF+uuvz4knnsgDDzzQOQLEfvvtR48ePZg5cyb77bdf2dtw9913s88++3Tev/POO+nRowe77bZbyuW32GILRo8ezWuvvca5555b9vaF5do3LS0trFmzhnvuuadLWcWdd96ZdRv7778/l19+OS+//HLaTHKvXr1yGulj9913p6mpiTvvvJMJEyZ0Tr/rrrtYvXp13hcoSW7DPvvsw9KlSznkkEN49913FSCLVAEFyCJSU1566SXmzp3bbfrYsWPzumDIIYccwvbbb89OO+3E4MGD+ec//8lf/vIXTjnlFAA22WQTzjnnHE4//XTeeustxo0bR3NzMx988AGPPPIIJ554InvvvXfJ9uuBBx7g7LPPZv/99+fZZ5/lkksu4bjjjmOzzTZLubyZccMNN3DIIYewcuVKjjjiCIYOHcrs2bN5+umnGTVqVLcLaJRKrn2z3377seeee3LKKacwd+7czlEschmG7swzz+Q3v/kN++67L1OmTGHbbbdl7ty53H///dx0003079+frbfemvnz5/PTn/6UsWPH0tzc3FkvHLbeeuvxve99j2nTptG3b18mTpzIG2+8wZQpU9hzzz271atnc9NNN/HEE08wceJERo4cydy5c5k2bRobbrghY8aMyWtdIhINBcgiUlO+8pWvpJw+Z86cvDJ3n//855kxYwY33HADy5cvZ9SoUXz/+9/nggsu6FzmyiuvZKuttuKGG27ghhtuwMwYOXIkEyZMSBu4Fur222/nmmuu4ac//SlNTU2cdNJJ/OhHP8r4mIkTJ/LEE09wxRVXcOKJJ9Le3s7666/P7rvvzpFHHlnS9iXLtW/uvfdezjjjDM477zwaGhr44he/yE9+8hMOPfTQjOsfNGgQTz31FFOmTOGqq65i3rx5DB8+nH322aezfOTEE0/kmWee4fzzz2fhwoVsvPHGvPfeeynXd8UVVzBs2DBuuukmbrzxRoYMGcJxxx3HtGnTug0RmM3222/Pgw8+yHnnncenn37Keuutx5577skdd9xB796981qXiETDXBFni0dt7Nix7vnnn4+6GSJV6Y033mCrrbaKuhkiEqLXpUhhzOwF59zYUq1Pw7yJiIiIiIQoQBYRERERCVGALCIiIiISogBZRERERCREAbJIHavmk3RFao1ejyLxoQBZpE717Nkzp4soiEhltLe3p7yUt0gU2tpg2jT/tx5pHGSROvWZz3yGWbNmMWLECHr37t3tcsEiUhnOOdrb25k1axbDhw+PujkSE21t0NoK48dDS0vltz1hAqxcCU1N8OijlW9D1BQgi9SpAQMGAPDRRx+xatWqiFsjUt969uzJ8OHDO1+XUt+iDlBbW/2216zxf1tbFSCLSB0ZMGCAPpBFRGIm6gB1/HgfmCcC9PHjK7ftuFCALCIiIhIjUQeoLS0+a12yEo/774fXXy9ByypHl5oWERERiZkoa5BL6qOPYMSIsm/G6P+mc4tLdp12ZZBFREREYqalpcoD44SFC/3fIUPgpJNKuupZs+A3v4W1awD38OalXLcCZBEREREpj9Wr/d8NN/TjxpXQrdPgQgdrHMDDJR2KSeMgi4iIiEh5JEZJaix9TjZRq93QAEBJa4YVIIuIiIhIeSQyyKEAuVQXIUmcTHjZZQBv/7u4tXWlEgsRERERKY9EgBxcJbLUYzwnarXPP3/JshK0tpMyyCIiIiJSHkklFqnGeI4jBcgiIiIiUh5JJRbhuuE4X4REJRYiIiIxUDPj3oqEJZVYlPwiJGWiAFlERCRipa7LFImNFKNYVMMYzyqxEBERiVi11GWK5C3FKBbVQAGyiIhIxKqlLlMkb1UaIFdXa0VERGpQtdRliuQtUWIR1CBXCwXIIiIiMVANdZlS+0p+sqgyyCIiIiJSlVatYul2Lezy5kvsEkxyDWC5Pn7nneHppzuv+9ypSgPkitYgm9npZva8mXWY2fSkeRPM7E0zW25mj5nZxpVsm4iIiEjdmjmTfm++QCNrOm+2Zo0/czSX27PPwscfd19vlZZYVPokvY+Ay4FfhSea2VDgXuBCYD3geeCuCrdNRESkLrS1wbRp/q8IAB0dALxpW9Krxyr6N6+i7YlVPsDNdts4yGkmguGwKs0gV7S1zrl7AcxsLLBRaNaXgdecczOC+VOBuWa2pXPuzUq2UUREpJZpzOXqV5aLyqxYAcBGm/Vm6uTG/Nbd1OT/KkAuuW2AlxN3nHPLzOy/wfQuAbKZnQycDDBq1KhKtlFERKTqpRpzWQFy9SjbF5wgQO63Xi/OOy/PxybKJ1au7D4v6Up61SIu4yD3AxYlTVsE9E9e0Dl3s3NurHNu7LBhwyrSOBERkVqhMZerW9kuKhOUWNDcnP9jM2WQU1xJrxrEpbVLgQFJ0wYASyJoi4iISM3SmMvVLfEFJ5FBLtkXnCCDXFCAnMgOq8Si5F4Djk/cMbO+wCbBdBERESkhjblcvcr2BScRIPfqlf9ja7DEoqIBspk1BttsABrMrBlYDfwe+KGZHQb8GbgIeEUn6ImIiIh0VZYvOCqx6KLSNchTgHbgXOCY4P8pzrk5wGHAFcACYDfgqAq3TUREpCI0zJrETikyyCqxKIxzbiowNc28vwJbVrI9IiIilaZh1iSWSlGDXMYSi7IMbZdBdYXzIiIiVa7WhlmrdOBS0z7+GBYvzu8x/frBiBHFbzvGJRZRfKlUgCwiIlJBZRuFIALKhpfQQw/BgQcW9ti77oIjjihu+zEusYjiS6UCZBERkQqqpWHWai0bHqlXX/V/Bw2Cz3wmt8fMm+dvL79cugA5hiUWUXypVIAsIiJSYbUyzFotZcMjFwSXbdudAldd1Xl8ZCxhufZaOOssWLas+O3HuMQiii+VCpBFRESkILWUDY/aB/9dyUjg0b83ceUE36+QpYSlb1//d/ny4hsQ4xILqPyXSgXIIiIiUrBayYZHbWYQILe7Xl0uIZ2xhKVPH/+3FBnkGJdYREEBsoiIiEjEPruhDy5XW1OXcpWMJSylzCDnWWLRpfSjBi8UUl2tFREREalBGw71AfK+E5s49IJ1meKMJSzlyCDnUGKRPHrJ21/uyQjQhUJEREREpISC8oT9JjVBKBBOV8LS1gb/ubcvx0LFM8jJo5e8NysIkGuoxKLSl5oWERERkWSJ4DJRrpBBIoN7/S98BnnZp5XNICdGL2lo8H9HfS7DSXoqsRARERGRguQRICcyuIvX+hrkFQuW0zdpmbyvcJjHSXrJo5eMfCxDDbJKLERERESkIHkEyIkMbkdHH1gL/axrBrmgKxzmeZJel9KPJ2tvFAuVWIiIiNSotjaYNs3/LcVyUkZ5BMiJDO53zvd5416ru9Ygp7rCYVblGgdZJRYiIiISF7lmEQvKNkrp5REgQ5DB3bEPXE63USwKusJhMeMgZxrmrUpLLJRBFhERqUG5ZhELyjZK6eUZIAM+29ujh39sIhBlXYb5ssvy+MJTzKWmdaEQERERib25czlw9FzuaoRVDno2woGjgTe7L3rgaHJaLqsNN4QBA4ppdX0rJEA282MhL13qh3oL9X/eVzhUiUUX1dVaERERyeytt2DMGHZcvZqXEtM6gK+lXnxHyGm5rAYNgpkzoX//AldQ5woJkMFfTW/pUl9mUcwXFJVYdFFdrRUREZHM3nzTByV9+8JGG1Vmm++8AwsXwocfwlZbVWabtaaYABmKv1iISiy6UIAsIiJSSxJBygEHwD33VGabO+4IL70E7e2V2V4tKjRALtXlplVi0YVO0hMREakliYAk30CrGImsYyLIkvxFmUFeu3bd9gsJkGuwxEIBsoiISC1JBCmV/ElbAXLxSpBBLng86/C2zfJ8MCqxEBERkZiLIkDu3dv/VYlF4YrMIL/54nImXFzgeNbFnKAHKrEQERGRmFMGuToFAfJ1NzbllwEOMshvvrCs8PGsiw2QVWIhIiIisaYAuSqtXu5Hkbj0qiYmTMijTCLIII/ZZDlNTdDQkMfV8xISI1gUUn8MKrEQERGRmEsEKSqxqC7B89a+tolVQQY4pxKJIIO86frLePRR/7jx4wu8SIhKLDpVV2tFREQkM41iUX2co2GND5BXW1N+GeDQKBZ5Xz0vQSUW3ajEQkREpJaoxKLqtD25BnOONfSAhgauuy6PQLcU4yCXq8Ri7Vpwzo+M0dBQePsioABZRESkzAoefqsQGsWi6jz5Nx9YrqQJ52DevDweXIpxkMtVYlGl5RWgEgsREalDbW0F1moWuK0JEwocfqsQyiBXnXEtPkDuoFf+J9iVIoNcrhKLKi2vAAXIIiJSZyodsLa20m34LQXIErbrDj5A7tm3iUcfyfP4KEUGuVwlFlU6ggWoxEJEROpMqoC1nMaPp/DhtwqhEovqEwSWfQc15f/lKQ4ZZJVYiIiIVLdEwNrR4c8dGjKkvNtraaHw4bcKEcUwb8ogF6fQq+hBaWuQC80gq8RCRESkurW0wHXXwemn+yzyd78L225b3sC14OG3QnKum9Ywb9WnmAA5kUH+xz9g4sTCtv/BB/5vsRnkGiqxUIAsIiJ1Z948PwLV2rUVqgsuUl510yqxqD7FBMijR/ufQubPhwcfLK4do0cX9rhEhnjNmnXDuoFKLEREREqirQ1uuMF/0JbRSXNhE2Ct+ZNx9vkb8EpZN1mUga/Dr1bAHDeUizquoLV1QLwCZGWQi1NsgPyvf8H77xfXhuZm2Guvwh5r5oPg1av98ZfYD5VYiIiIlMBllxWfBcvBUOCIxJ01wF/LvsmibB3cAF5oaGH8+K+lX1gBckWH8SuJYgJkgG228bcoNTX5gDjxMweoxEJERKQkFi/2fy++GLbYItq2xMyia37BwBf+xoVnLmWTXGqQKxiU/Os/vdkWWDKnnf4V22pqFR93uhSKDZDjINVIFiqxEBERKYFEDesXvgBjx0bblpgZ+NRT8MLf2GRElixtBUexaGuDW2+FF3/ZzD+Ad19fwbK2aAPSio87XQq1GiCrxEJERKQEEgFy4qQvWSdRxpC4qEM6FRrFIpGpXbECtnC+bU1uBX9ujTYgTQzjl8ggl33c6VKohQA50fbwSBYKkEVEREpAAXJ6iTFqcw2Qy5xBTmRqnYN2/PPVx9ojD0grPu50KdRCgJypxEI1yCIiIkVQgJxezALkcKZ2bY9mWAXrD1rBqBgEpKUYd7qiajVAVgZZRESkBJIC5KobjaCcEgFytpEiKhQghzO1E3ZuhgOgaU08RrGoOrUQIKvEQkREpEwSwV9zc3WORlBOQQ3y8091sCrTiXAV/Fm7M1PboQuFFKUWAuQaK7HoEXUDREREgHXDDphBr14pRyOoZ+/M8hnkF9s62Htvn11PKYqgJBHYrVpV9ou81KRaDZCrOIOsAFlEROIhlD3GrLPGtaHB32bOzBAU1oEnn/MBchMddHT44dVSquAwb53Mch9lQ7qrhQC5xkosFCCLiEg8JNUfJ2pcTzrJx18//7kvuajXIHl1gw+QexGPYd666a0yi4IlvlRUc4CcKYOsEgsRkfhpa4Np0+o3sKoaKUawaGmBUaP852y9l1r8zwSfoW1mBU1NcNxxaRaMqu4zZpebriq1kEHWlfTKx8xGAzcCLUAH8Dvgu8651VG2S0Sql070qiJphnirygs/lMEW2/kM8nabd9A6PR4n6XURYYBc9aOd1EKAXGMlFnFr8Y3Ap8AGwCDgEeA04PoI2yQiVawqLztbr8I1yCFVeeGHcgiGedtkow42ydQHUQXIEZVY1MSX4DIEyLl+aSjZl4saK7GIW4D8WeAnzrkVwCdm9hdgm4jbJCJVTNnHKpLhIiHluvBDVWUeY3ahkG4iyiDXxJfgEgfIuX5pKOmXixorsYhbDfJ1wFFm1sfMRgAHAX8JL2BmJ5vZ82b2/Jw5c6Joo4hUkUT28bLLqjSzVE/KfBW95Fr0RHBw4YVVcvJfrqNERDGKBUQWIIdHO6naL8GJ5yzxJahIuQ6RWNKhFJNKLNra4KE/q8SiVJ4ATgYWAw3Ar4H7wgs4524GbgYYO3asq3D7RKQKVd1lZyMWWVa1jAFyqkxZ1WUec7mSnnN1V2JREyU4Jc4g5/rLWUl/YUscb9/8Jqu+fRabLIQ+LAfg47k92aCIVUchNgGymfXAZ4tvBvYA+gG/Aq4Gvh9h00RE6kak9ZxlDJBTBcNVV36TS4lF4iIdPXr4lGolRXiSXtV/CS5xgJzrl4aSfrnYfXf49a9h6VJ6spTPBJPXYjxnu/LFIlYdhdgEyMB6wCh8DXIH0GFmtwCXowBZRKQiSpVVLSgLXcYAOVUwXHWZx1wC5Cgv7ath3gpXhpP0cv3SULIvF6eeCkcdBStX8txzcPjhfrdcUy9+f+bAEmygsmITIDvn5prZu8A3zexH+Azy8cAr0bZMRKR+lCKrWnAWOs0oFqWQLhiuqsxjLjXIUQbIulBI4WphmDeAQYMA2GUS3Pm3KvrymUJsAuTAl/En6p0DrAH+BpwZZYNEROpJKbKqBWehy3ySXlUFw6nkUoMchwzyww/D8uWV3341e/tt/7faA+SQan+9xSpAds69BIyPuBkiInWt2A+2grPQZQ6Qq17cSyyC7CF33OFvkr/+/aNugQRiFSCLiEj1KzgLrQA5s3CA7ByYdV8mqiHeAE4/3V8YQiUWhRkxAvbcM+pWSEABsoiIlFxBWWgFyJk1NPjxZFev9rdUQXAigxzFT/UbbwzXXlv57YqUQdwuFCIiIvVKAXJ22cosoiyxCCRfkEWkGimDLCIi8VDGUSxqRq9esGyZ76t+/brPjzhAjnQcbZESUgZZRETiQRnk7GKeQS7ppYtFIqQAWURE4kEBcnYxD5ATI5g0NFTJ1QlF0lCJhYiIxIMC5OyyXSwkylEsqMKrE4qkoQBZRETiQQFydtkuFhLlKBaBar9AhAioxEJEROJCAXJ2MS+xEKkVCpBFpKQqMcSThpGqUTU6ikVJj1cFyCIVoRILESmZSgzxpGGkalgNZpBLfrxmq0FWgCxSEsogi0jJVGKIJw0jVcNqMEAu+fFaxxnkavzlqBrbLJ4yyCJSMokhnhLZsnIM8VSJbUhEajBALvnxmutJejUWIFfjL0fV2GZZRwGyiJRMJYZ4itswUm1t8WlL1avBALnkx2u2DHJimLcIR7Eoh1SZ+Li/3qqxzbKOAmQRKalKDPEUl2Gk2trgwH1WMmblizzSuIbrr4cxY6JuVWqvvgov/hN22jG+bazFABlKfLzWaQ1yNf5yVI1tlnUUIIuIFKi1FX7ccQqT3XRYCZwacYMyGBPcYi9xCTZJrU5rkOP2y1EuqrHNso4CZBGRAo0fD028CsArth2jx/RjQP9o25TKhx/CzJngAANGjYKNNoq6VWlMnAhmUbcivuq0Bhni88tRPqqxzeIpQBYRKVBLC6zYcAHMgjW/ncGAIzePukkpfdAG+4ZPFroTNtKHdnWq0wyySKUpQK5SOjFIJB6a2xcAsOM+g8u6nWJe8/qpt4bUaQ2ySKUpQK5CGjpGJCbWroUFPkBmcPkC5FK85vVTb43IdRQLBcgiRcn5QiFmdp+ZfcHMdHGRiOlCCSIxsXgxOAf9+0Nj+fINes1Lp1CAnPIiFIkMsk50FClKPu/oy4C7gEVmNh24xTn3dllaJRlp6BiRmKhA9hj0mpeQIED+5L0VqX9VUImFSEnknA12zh0NbABcBuwLvGVmT5jZcWZWW4NWxlyinvCyy1ReIRKp+fP93/XWK+tmon7N63K5MRLUIL/0bAcdHSl+VVCALFISef0m6JxbDPwU+KmZbQOcCPwMuN7M7gKuc869UfpmSjLVE4rEQIUyyBDNa76tDW69FX71Kx+I6ZyH6L09sxebARNm3co87vET10C/K4GrgeXL/TQFyCWhE+LrV0FFc2a2IXAI8AVgNXAPMBJ4xczOc879qHRNFBGJqQoGyJWWODFwxQpfZg26XG4cPLZoJ9anH/1ZyiAWrZuxNLRQczOMHVvxttUanRBf3/I5Sa+nmR1uZg8A7wOHAj8ANnDOneCcmwgcBkwpS0tFROKmQiUWUUicGJgIjs1U/xwH2x65NSOb5zKkxwI2aF7A9Gv938T9Zx9a4L+4KZIrmk6OrW/5ZJA/xl+E6TfAuc65V1Is8wSwoBQNqwb66UWkztVwBjl8YmBjI3z963DccXqvi1pLCzz4t160tvZi/Hj/GTRnFaxZCw2r4NEXYNf9o25lbdDJsfUtnwD5TGCGcy7N9S3BObcQ+GyxjYqVG2+E//1fWL26y+S1DnZc3YM/2lQmNJ+rn15E6lEiQK7BDLIuLhJfyfXoCuLKQ6+B+pZzgOycu62cDYmtP/0J2tu7Te4BNAMHu/v5wcpzVZcnUo9qOIMMOhm4GiiIKy+9BuqXLvqRTWLInD//2V+5KLi98otnAehNu761i9SrRA1yjQbI5abh40qjpQXOO0+BnEgp6VLT2SQC5N69u1yZaLv/6Q/AyKErePQPemMSqUs1XGJRbhohQETiTBnkbNINuh4M1j6kT7ve1EVirmyZyhovsSgnjRAgInGmDHI26QLk3sHFA1ekPWdRRGIgnKks+WgMKrEomEYIEJE4U4CcTZYMcqoT+ETq0tq18O9/+78x8q+7YJMOPwwWa+CJm+Aft/irw+2wQ5Err+FxkMtNJ5eJSJwpQM5m5Ur/VxlkkcxOOAGmT4+6Fd2cHNy66ACOLtEGevSAAQNKtLL6ohECRCSuFCBnky6D3LOnv7TUqlW+iK6hofJtE4mTV1/1f0ePXvcFMiaWt8OiRbBwATj8S3fjjaFPKZp58ME+SBYRkZqhADmbdAGymQ8Cli/3WeS+fSvfNpE4Sfzacu+9sOOO0bYlSZ/g9l7o6pdbKXMpIiJpKEDOJl2ADL4OeflyX4esAFnqXeK1EhoOMW70k76IiORCvwtmkylAVh2yyDqZXisiIiJVRAFyNtkyyKCRLCTWKna1MgXIIiJSI1RikY0yyFLFKnq1MgXIIiJSI5RBzkYZZKliFb1amQJkERGpEQqQs8klg6wAWWIqcbWyhoYKXK1MAbKIiNQIlVhksnbtuquCNaboqkQGWSUWElMVvVqZAmQREakRCpAzSXzgNzb6cY+TKYMsVaBiQ5spQBYRkRqhEotMsn3gK4MsVaIiI1koQBYRkRqhDHIm2T7wlUHOSVtbhX7il5QqMpLFmjXgnL/ksi67LiIiVS52GWQzO8rM3jCzZWb2XzPbK7LGKINctERwduGF/m/Zx+KVbrKNZFGS7HLiMtPKHouISA2IVQbZzPYDrgaOBJ4FNoi0QcogFy1VcKYscmUlRrJIZJDDI1mULLus8goREakhsQqQgUuAS51zzwT3Z0XZGGWQi5cpOJPKyDSSRcm+wChAFhGRGhKbANnMGoCxwB/M7D9AM3AfcLZzrj203MnAyQCjRo0qb6OUQS5aRYcZk7TSjWRRsi8wdRggq7ZeRKR2xSZABoYDPYHDgb2AVcD9wBTggsRCzrmbgZsBxo4d68raImWQS6Jiw4xJ3kr2BSZmAXK5g9eKXsJbREQqLk4BciIN+/+ccx8DmNn/kRQgV5QyyFIHSvIFJkYBciWCV9XWi4jUttiMYuGcWwB8CISzwuXNEGejDLJIbiIIkNONvpFt1I5SqOglvEVEpOLilEEGuAX4tpn9BV9icSbwp8haowxy2eTyE7hqPKtIhQPkTFniSpwYqtp6EZHaFrcA+TJgKPBvYAVwN3BFZK1RBrkscvkJXDWeVabCAXKmEodKBa+qrRcRqV2xCpCdc6uA04Jb9JRBLotc6jdV41llEq+VpqaKbC5blljBq4iIFCNWAXLsKINcFrn8BK7xk6tMhTPIKnEQEZFyUoCciTLIZZFLcKMAqMpEcKlpZYlFRKRc6iJALvhkL2WQyyaX4EYBUBWJ0TBvIiIixar5ALmok72UQa5rGkUjDwqQRUSkhtR8gFzUyV65ZpAVINccjaKRJwXIIiJSQ2JzoZByKWpA/1wzyCqxqDmVuNhETVGALCIiNaTmM8hFneyVbegqZZC7Wr0aXnnFR5VVbtJn4IFGWOWgZ6O/z3NRtyoeXnkFXnwRtv7ipux6wGA/UQGyiIjUkJoPkKGIk72qIIMcqzrZb34TfvGLiBtRGtsBf0/c6QBOjK4tcbNdcJt943CeeeIDdt+rpwJkERGpKXURIBcs24d+UxOY+eX+8x/oUdmKlRdfhBOO8Zu/rSfcfjvstBMwZAgMHFjRtgDw6qv+71ZbQd++ld9+lVi6FJYsgf79oV+/qFuTn48/hg9nwY78k+HM5rd/Wcjuew2r+IVCREREykkBcibZAmQzX2bR3g6bbVa5dgV2Al5P3OkAvhL837s3vPUWjBxZ2QYtXuz/3n03jBlT2W1XiS4n/82vvpP/3gva/1b7SEbyIXuNDcqLlEEWEZEaogA5k1w+9L/9bZgxozLtSbKiw2f0nPOx+gYbQPOCT3zA/tpreQXIJSnVSATIAwYUuILaV+2X0E7U9Pc9uDfMg523Wu5nKEAWEZEaogA5k1w+9K++2t8i0Ax8EgpsP9sCHHmkz+AuXJjzeko2pJkC5Kxq4RLaLS3ARn1gHutOUFWALCIiNUQBciZV8KHf7QTERO3xokU5r6MkWc21a31hLfjiWkmpZi6hnThBdXmQQY7gUtMiIiLlogA5kyoIkLspIEBOl9XMq+xi2TJf69G3rx90WtKqiUto9+nj/yqDLCIiNUgBcibV+KE/aJD/m0eJRaqsZt5lFyqvKItYDeMXlpxBrsbXioiISBoKkDOpxg/9AjLI0D2rmXfZhQLkkov15a5rMIMc2y8jIiJScQqQM6nGD/0CA+RkeZ9Mlqg/jlGAXO0BT6xHvKixDHKsv4yIiEjFKUDOpBo/9BMBch4lFqnkfTJZzDLItRDwxHrEixrLIMf6y4iIiFScAuRMqvFDP1GDXGQGGdKfTJYyMxuzALkWAp5Yj3iRyCAnB8hVeiW9WH8ZERGRilOAnEmMAuScywVKVGKRqR0pM7OJADkmQ7zVSsAT2xEvEhnkGimxiPWXERERqTgFyJnE5EM/r3KBEpVYpJM2MxtxBjn5C4QCnjJLl0Gu0gAZYvxlREREKk4BciYx+dDPq1ygzBnktJnZCAPkdF8gFPCUUY1lkEVERMJ6RN2ASmlrg2nT/N+cxeRDPxGUNjTkUC6QCFCXLPFXtyuxRGb2ssuSMtkRBsipvkBImdVgBllERCShLjLIBY9oEJMP/bzKBRoafB3wkiWwZAltrw8seZlBysxshAFyrdQbV5XkDLIuNS0iIjWkLgLkgkc0iEmADHmWCwwcCEuW8MKjC5lwzMCChzrLaxzhCANk1RtHQBlkERGpYXURIBecYazWD/2BA+HDD/ln66KChzqrtktNq964wlSDLCIiNawuAuSCM4zV+qEfjIW8+1aLCi490KWmJSNlkEVEpIbVRYAMBWYYq/VDPxjJYsxGCwsuPcg7664Aub6kyyBX6YVCREREwuomQC5IlQfILFpEy8GFlR7knXVfssT/zRIg51XXLPGlDLKIiNSwug+QMwZs1fqhX6KxkPPKuueQQS54NBGJH9Ugi4hIDavqAHnZMj+2caHZyKwBW7V+6Ac1yNx/PyxYUJltJq7clyFALng0EYkfZZBFRKSGVXWA/NZbcOGFhWcjswZs1fqhv/76/u8jj/hbpQwcCL16pZ2t8YpriDLIIiJSw6o6QHauuGxk1oCtWj/0J0+Gjo51ZQ+VsvfeGWdrvOIaogyyiIjUsKoOkM2gR4/cs5E33wz33AOHHQYnn5xDwFatH/oDB8L3vx91K1LSeMU1ornZ/+3o8N9Sq/W1IiIikkJVB8hbbAHHHZdbNvLmm+GUU/z/Dz/s/yaC5LSP1Ye+SGpmPovc3g4rVuhS0yIiUlN6RN2AYvTtC+edl1tG8p57Mt9PSQGySHqJOuT2dr1WRESkplR1BpnXXoOttspp0TsXwieh++u/AmR76IoV/q8+9EW6S9QhL1+uC4WIiEhNqe4AecUKePPNnBYdHNw6fULXiDmdzTeHxuruJpGyUAZZRERqVHVHfttsAzNmlHcbo0f7eksR6SpVBlkBsoiI1IDqDpCbm3MusRCREktkkJctg7Vr/RfJhoZo2yQiIlICVX2SnohEKJFBToy3reyxiIjUCAXIIlKQ+R0+g/zx9If8BAXIIiJSIxQgi1S5tjaYNs3/reQ2H35mAAAbzLjeT0yUXIiIiFS56q5BFqlzbW0wYcK6y6U/+mhlrlTY2gq/c2ezkgZ62Uq2HQNbn3dI+TcsIiJSAQqQS6CtLcPlqkXKqLXVB8dr1vi/ra2VOQbHj4fLeu3AN1be6gPznwE69kVEpEbUdYBcisA2qgyeCPhjt6lp3fE3fnxlttvS4o91fTEUEZFaVLcBcqkC26gyeCIQbaDa0qJjXUREalMsA2Qz2wz4F/A759wx5dhGqQLbqDJ4IgkKVEVEREorlgEycAPwXDk3UKrAttZ/alZ9tYiIiNSb2AXIZnYUsBB4Gtg02/KFBnClDGxrNYOn+moRERGpR7EKkM1sAHApsA9wYpplTgZOBhg+fJOiArhaDWxLRfXVIiIiUo/idqGQy4BfOuc+TLeAc+5m59xY59zYxsZB3QI4KZ1EGUpDg+qrRUREpH7EJoNsZjsA+wI75vqY/v1h/nydIFcutV5fLSIiIpJKbAJkYDwwGphpZgD9gAYz29o5t1OqB/TtqwCu3FSGIiIiIvUmTgHyzcCdofv/iw+Yv5npQQrgRERERKSUYhMgO+eWA8sT981sKbDCOTcnulaJiIiISL2JTYCczDk3Neo2iIiIiEj9idsoFiIiIiIikVKALCIiIiISogBZRERERCREAbKIiIiISIgCZBERERGREAXIIiIiIiIhCpBFREREREIUIIuIiIiIhChAFhEREREJUYAsIiIiIhKiAFlEREREJEQBsoiIiIhIiAJkEREREZEQBch1rK0Npk3zf0VERETEa4y6ARKNtjaYMAFWroSmJnj0UWhpibpVIiIiItFTBpn6zKS2tvrgeM0a/7e1NeoWiYiIiMRD3WeQ6zWTOn6839/Efo8fH3WLREREROKh7gPkVJnUegiQW1r8l4HWVh8c18M+i4iIiOSi7gPkes6ktrQoMBYRERFJVvcBsjKpIiIiIhJW9wEyKJMqIiIiIutoFAsRERERkRAFyCIiIiIiIQqQRURERERCFCCLiIiIiIQoQBYRERERCVGALCIiIiISogBZRERERCREAbKIiIiISIgCZBERERGREAXIIiIiIiIhCpBFREREREIUIIuIiIiIhChAFhEREREJUYAsIiIiIhKiAFlEREREJEQBsoiIiIhIiAJkEREREZEQBcgiIiIiIiEKkEVEREREQhQgi4iIiIiEKEAWEREREQlRgCwiIiIiEqIAWUREREQkRAGyiIiIiEiIAmQRERERkRAFyCIiIiIiIbEJkM2sl5n90szeN7MlZvaSmR0UdbtEREREpL7EJkAGGoEPgHHAQGAKcLeZjY6yUSIiIiJSXxqjbkCCc24ZMDU06U9m9i6wM/BeFG0SERERkfoTmwA5mZkNBzYHXkuafjJwcnC3w8xerXTbasxQYG7Ujahy6sPiqP+Kpz4snvqwOOq/4qkPi7NFKVdmzrlSrq8kzKwn8CDwX+fcKRmWe945N7ZyLas96sPiqQ+Lo/4rnvqweOrD4qj/iqc+LE6p+y9ONcgAmFkP4DZgJXB6xM0RERERkToTqxILMzPgl8BwYKJzblXETRIRERGROhOrABn4KbAVsK9zrj2H5W8uc3vqgfqweOrD4qj/iqc+LJ76sDjqv+KpD4tT0v6LTQ2ymW2MH62iA1gdmnWKc+6OSBolIiIiInUnNgGyiIiIiEgcxO4kPRERERGRKClAFhEREREJiV2AbGbrmdnvzWyZmb1vZl9Ls5yZ2dVmNi+4XR2MgpGYv4OZvWBmy4O/O1RsJyKWRx+ebWavmtkSM3vXzM5Omv+embWb2dLg9nBl9iBaefTfVDNbFeqfpWb2udB8HYPZ+/DBpP5baWb/Cs2v12PwdDN73sw6zGx6lmXPNLNPzGyxmf3KzHqF5o02s8eCY/BNM9u37I2PiVz70MyOD16fi83sQzP7gZk1hua3mtmK0DH4VkV2IGJ59N9kM1uT9DoeH5qvYzB7H96U1H8dZrYkNL9ej8FeZvbL4DNkiZm9ZGYHZVi+pO+FsQuQgRvwYyAPB44Gfmpm26RY7mTgUGB7YDvgYOAUADNrAu4HbgcGA78G7g+m14Nc+9CA4/B9dCBwupkdlbTMwc65fsFt/3I2OkZy7T+Au0L908859w7oGCTHPnTOHRTuP+BpYEbSYvV4DH4EXA78KtNCZnYAcC4wAdgY+BxwSWiR3wL/BIYAFwC/M7Nh5WhwDOXUh0Af4Lv4q5jthu/L/01a5vTQMVjSq3XFWK79B9CW9D7YGpqnYzBLHzrnTk16H/wt3d8H6/EYbAQ+AMYBA4EpwN1mNjp5wbK8FzrnYnMD+uI/VDcPTbsNuCrFsk8DJ4funwA8E/y/PzCL4CTEYNpM4MCo9zFOfZjisdcD/y90/z38kHuR71cc+w+YCtyeZj06BvM8BoHRwBpgdGha3R2DSX1yOTA9w/zfAFeG7k8APgn+3xw/KlD/0Py/A6dGvV9x6sMUy58F/DF0vxU4Mer9iGv/AZOBJ9PM0zGYQx8mLdsXWAKMC02r62MwqX9eAQ5LMb3k74VxyyBvDqx2zv07NO1lIFX2bptgXqrltgFecUEvBF5Js55ak08fdjIzA/YCXkuadYeZzTGzh81s+9I2NZby7b+DzWy+mb1mZt8MTdcxmOcxiP814+/OufeSptfbMZiPVO+Dw81sSDDvHefckqT59XAMFuPzdH8fnGZmc83sqXD5gHTaMeiff5vZhaESFR2D+TsMmAM8kTS97o9BMxuO/3xJfn1CGd4L4xYg9wMWJ01bBPRPs+yipOX6BYFe8rxM66k1+fRh2FT88XBLaNrR+KzexsBjwENmNqgUjYyxfPrvbvyFbYYBJwEXmdlXQ+vRMbhOLvt+HDA9aVo9HoP5SPU+CL6v6/kYLIiZfQMYC/woNPkc/M+1I/AXIvijmW0SQfPi6glgDPAZfHD3VSBxPouOwfwdD9yalFyp+2PQzHoCdwC/ds69mWKRkr8Xxi1AXgoMSJo2AP9zQ7ZlBwBLg4Mqn/XUmrz33cxOxwcnk5xzHYnpzrmnnHPtzrnlzrlpwEJ8lrmW5dx/zrnXnXMfOefWOOeeBn4MHJ7vempQIcfgnsD6wO/C0+v0GMxHqvdB8H1dz8dg3szsUGAacJBzbm5iunPuH865Jc65Dufcr4GngIkRNTN2nHPvOOfedc6tdc79C7gUvQ8WxMxGAeOBW8PT6/0YNLMe+DK9lcDpaRYr+Xth3ALkfwONZrZZaNr2pE6nvxbMS7Xca8B2QTY5Ybs066k1+fRhImNyLjDBOfdhlnU7/Il9tSyv/ksS7h8dg/n14fHAvc65pVnWXQ/HYD5SvQ/Ods7NC+Z9zsz6J82vh2MwL2Z2IPBz/Amh/8qyuI7BzJLfB3UM5u5Y4CkXnOydQd0cg8Fn6C/xJ3wf5pxblWbR0r8XRl1wnaLQ+k782YZ9gf/Bp8G3SbHcqcAb+J8cNgx29NRgXhPwPvAdoBf+G8f7QFPU+xezPjwa+ATYKsW8UcFjm4Bm/E9mc4AhUe9fjPrvEPwIFQbsij8p73gdg7n3YbBs72D+PjoGO/e9MdjnafjMSTPQmGK5A4PX8NbAIOBvhE6GBJ7Blws0A1/CZ+CHRb1/MevDfYB5wOdTzBsEHJB4bPCeuYzQCai1esuj/w4Chgf/bwm8Clwcmq9jMEsfhpZ/C/hG0rS6PQaD/b8pOIb6ZVmu5O+Fke98ip1cD7gvOABmAl8Lpu+FL6FILGfAD4D5we0HdB0xYEfgBaAdeBHYMep9i2Efvguswv/8kLjdFMzbBn9S2bLgw+NRYGzU+xaz/vtt0DdLgTeBM5LWo2MwSx8G076K//JgSdPr+Ricis8ShW9T8V8algKjQsueBczG133fAvQKzRuNPwO+Hf/hWzcjguTah/ja9tVJ74MPBvOGAc/hf4pdiP+Q3S/qfYtZ//0oOP6WAe/gSyx66hjM+3XcEvRh/6R11PMxuHHQZyuSXp9HV+K90IIHioiIiIgI8atBFhERERGJlAJkEREREZEQBcgiIiIiIiEKkEVEREREQhQgi4iIiIiEKEAWEREREQlRgCwiIiIiEqIAWUREREQkRAGyiIiIiEiIAmQRkSpjZsPM7GMzuzg0bTszW2FmX4mybSIitUCXmhYRqUJmdgDwR2Ac8BLwPPCsc+7rUbZLRKQWKEAWEalSZnYd8EXgcWAvYAfn3NJIGyUiUgMUIIuIVCkz6wW8DGwG7OGc+0fETRIRqQmqQRYRqV6jgZGAAz4XbVNERGqHMsgiIlXIzHoCzwD/Bv4BXAxs75ybGWnDRERqgAJkEZEqZGZXAV8DtgMWAQ8CzcA+zrm1UbZNRKTaqcRCRKTKmNk44HvAcc65hc5nOiYDWwPnRNk2EZFaoAyyiIiIiEiIMsgiIiIiIiEKkEVEREREQhQgi4iIiIiEKEAWEREREQlRgCwiIiIiEqIAWUREREQkRAGyiIiIiEiIAmQRERERkZD/D9YDkpRzpTKhAAAAAElFTkSuQmCC ", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "### ENTER CODE HERE ### ", "fig = plt.figure(figsize=(10, 5)) ", "plot_predictions(dt_reg, X_train, y_train, [0.0, 2.0, 2.0, 12.0], label='Ensemble Predictions') ", "fig_id = 'best_dt_2DLinearReg' ", "plt.xlabel('x') ", "plt.ylabel('y') ", "plt.title('DecisionTreeRegressor: splitter="best", min_samples_split=10, max_depth=6') ", "save_fig(fig_id, tight_layout=True, fig_extension="png", resolution=300) ", "plt.show()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Compute Generalization Error ", " ", "Compute the generalization error for each of the optimal models computed above. Use MSE as the generalization error metric. Round your answers to four significant digits. Print the generalization error for all three models." ] }, { "cell_type": "code", "execution_count": 25, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Generalization Error of best GradientBoostingRegressor is 1.4088 ", "Generalization Error of best RandomForestRegressor is 1.2915 ", "Generalization Error of best DecisionTreeRegressor is 1.5705 " ] } ], "source": [ "### ENTER CODE HERE ### ", "from sklearn.metrics import mean_squared_error as MSE ", "y_gt = gt_reg.predict(X_test) ", "print("Generalization Error of best GradientBoostingRegressor is {:.4f}".format(MSE(y_test, y_gt))) ", "y_rf = rf_reg.predict(X_test) ", "print("Generalization Error of best RandomForestRegressor is {:.4f}".format(MSE(y_test, y_rf))) ", "y_dt = dt_reg.predict(X_test) ", "print("Generalization Error of best DecisionTreeRegressor is {:.4f}".format(MSE(y_test, y_dt)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Critical Analysis ", " ", "Think critically about the different algorithms, as well as their prediction results, and characterize the trends you observe in the prediction results. Can you explain in your own words how the algorithms work? How do the results predicted by these models compare and contrast? Can you see any relationship between the algorithms and the model results? Can you see any relationship between the prediction curves and the generalization error? ", " ", "I am looking for **meaningful content** here. Do not copy-and-paste model definitions off of the Internet. You should think and write critically. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "1. Decision trees work by breaking down the input space into progressively smaller chunks and building a tree out of it. Finally, decisions or target values are assigned to the leaf nodes. Random forest is an ensemble of several decision trees that are randomly constructed on the dataset. The predictions of all these trees are averaged to create the final prediction which is often more accurate than all of the individual predictions. Gradient boosting regression uses gradient descent to minimize a loss function. It uses several weak learners to map the input features to the residual values, which are the difference between the model predictions and the desired outputs. These residual values are added to the inputs to create a gradually converging and improving pattern of predictions over iterations. ", " ", "2. As seen from the experiments above, the random forest regressor gives the best results. It has more ability to generalize compared to other algorithms considered here. Decision tree regressor performs the worst. ", " ", "3. Yes. Decision tree regressor has a staircase like prediction over the input space. This aligns well with the way the algorithm works, that is, by assigning one particular value to certain chunks of input space. Random forest regressor averages several such randomly constructed decision trees over the input data, hence, its predictions seem more close to the underlying pattern of the data and less staircase like. ", " ", "4. Yes. Decision tree regressor is expected to give the worst generalization performance, as seen from its prediction curves. Hence its generalization error is also high. On the contrary, random forest regressor which is a random combination of several decision trees gives the best generalization performance. Its prediction curve fits the underlying pattern in the data. So its generalization error turns out to be the least of all as expected. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Final Model Selection and Justification ", " ", "Based on the arguments outlined in your critical analysis, make one final model recommendation. Which model best characterizes this data? In other words, which of these models is going to generalize better? Consider all factors. What is the type of the optimal model (i.e. `GradientBoostingRegressor`, `RandomForestRegressor`, or `DecisionTreeRegressor`)? What are the optimal model hyperparameters that should be used for training on this data set? Succinctly summarize the justification for your choice based on your arguments made above in the critical analysis section. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Based on the above observations, the random forest regressor is found to be the best model. It is expected to give the best generalization performance. The hyperparameters of the optimal model are: ", " ", "`splitter='best'`, ", " ", "`max_depth=6`, ", " ", "`min_samples_split=10` " ] } ], "metadata": { "kernelspec": { "display_name": "DL_gpu", "language": "python", "name": "dl_gpu" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.7" } }, "nbformat": 4, "nbformat_minor": 2 }

Assignment 2: Exploring Tree-Based Regression Methods for 2D Linear Data

Get Instant Homework HelpOn Your Mobile

Get Instant Homework HelpOn Your Mobile

Get Instant Homework HelpOn Your Mobile