{ "cells": [ { "cell_type": "markdown", "metadata": { "toc": true }, "source": [ "

Mis notas: Machine Learning con Scikit-Learn

\n", "
" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "> **Disclaimer**: Este notebook contiene mis notas sobre Machine Learning con Scikit-Learn, resumiendo básicamente el [capítulo 5](https://jakevdp.github.io/PythonDataScienceHandbook/05.00-machine-learning.html) de [Python Data Science Handbook](https://jakevdp.github.io/PythonDataScienceHandbook/index.html) escrito por [Jake VanderPlas](http://vanderplas.com/). Recomiendo leer la fuente original e ir ejecutando todos los ejemplos (___learn by doing!___).\n", "\n", "## Intro\n", "\n", "[Scikit-Learn](https://scikit-learn.org/) es un paquete para la implementación de Machine Learning en Python, construido sobre NumPy, Scipy y Matplotlib.\n", "\n", "La manera más apropiada de pensar en datos dentro de Scikit-learn es usando tablas. Una tabla básica de dos dimensiones contará con las distintas muestras (samples) del dataset en filas, y los atributos (features) como columnas. Es lo que se conoce como la matriz de características `X`, y suele estar contenida en un array de NumPy, un DataFrame de Pandas o una matriz de SciPy.\n", "\n", "Las muestras son observaciones de cualquier cosa que puede ser descrita por medidas cuantitativas de sus atributos o características.\n", "\n", "Aparte de la matriz de características, generalmente trabajaremos con un array unidimensional de etiquetas `y`, contenido en un array de NumPy o un objeto Series de Pandas, cuya longitud equivale al número de muestras en X (una etiqueta por muestra).\n", "\n", "Vemos un ejemplo con el dataset `iris` incluido con seaborn:" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
sepal_lengthsepal_widthpetal_lengthpetal_widthspecies
05.13.51.40.2setosa
14.93.01.40.2setosa
24.73.21.30.2setosa
34.63.11.50.2setosa
45.03.61.40.2setosa
\n", "
" ], "text/plain": [ " sepal_length sepal_width petal_length petal_width species\n", "0 5.1 3.5 1.4 0.2 setosa\n", "1 4.9 3.0 1.4 0.2 setosa\n", "2 4.7 3.2 1.3 0.2 setosa\n", "3 4.6 3.1 1.5 0.2 setosa\n", "4 5.0 3.6 1.4 0.2 setosa" ] }, "execution_count": 1, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import seaborn as sns\n", "iris = sns.load_dataset('iris')\n", "iris.head()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Construimos la matriz de características `X` y el vector de etiquetas `y` a partir del DataFrame original:" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
sepal_lengthsepal_widthpetal_lengthpetal_width
05.13.51.40.2
14.93.01.40.2
24.73.21.30.2
34.63.11.50.2
45.03.61.40.2
\n", "
" ], "text/plain": [ " sepal_length sepal_width petal_length petal_width\n", "0 5.1 3.5 1.4 0.2\n", "1 4.9 3.0 1.4 0.2\n", "2 4.7 3.2 1.3 0.2\n", "3 4.6 3.1 1.5 0.2\n", "4 5.0 3.6 1.4 0.2" ] }, "execution_count": 2, "metadata": {}, "output_type": "execute_result" } ], "source": [ "X_iris = iris.drop('species', axis=1)\n", "X_iris.head()" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0 setosa\n", "1 setosa\n", "2 setosa\n", "3 setosa\n", "4 setosa\n", "Name: species, dtype: object" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "y_iris = iris['species']\n", "y_iris.head()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Estimator API\n", "\n", "La API de Scikit-Learn está diseñada siguiendo unos principios:\n", " * **Consistencia**: todos los objetos comparten una interfaz común con un conjunto limitado de métodos\n", " * **Inspección**: todos los valores de parámetros se exponen como atributos públicos\n", " * **Jerarquía**: los algoritmos son (los únicos) representados por clases; los datasets por arrays de NumPy, dataframes de Pandas o matrices de SciPy, ....\n", " * **Composición**: Muchas tareas pueden ser expresadas como una secuencia de tareas más sencillas.\n", "\n", "Cada algoritmo de Machine Learning está implementado siguiendo estas líneas, componiendo una interfaz consistente. Para usar un algoritmo concreto seguiremos los siguientes **pasos**:\n", " 1. **Importar** la clase de Scikit-Learn con el modelo más adecuado\n", " 2. **Instanciar** la clase con los valores deseados para establecer los hiperparámetros del modelo\n", " 3. **Separar** los datos en una matriz de características y un vector de etiquetas. Es conveniente separar también en muestras de entrenamiento y muestras de validación.\n", " 4. **Entrenar** el modelo para que se ajuste a los datos usando el método `fit()`. Si tenemos un subconjunto de validación, lo siguiente es **evaluar** el modelo para ver si es conveniente modificarlo.\n", " 5. **Predecir** las etiquetas aplicando el modelo a nuevas muestras:\n", " * Para aprendizaje supervisado predecimos sus etiquetas usando el método `predict()`\n", " * Para aprendizaje no supervisado normalmente transformamos o inferimos propiedades de los datos usando `transform()` y `predict()`\n", "\n", "Vemos esto a través de ejemplos..." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Ejemplo de aprendizaje supervisado: Regresión lineal simple\n", "\n", "Usamos datos generados por nosotros mismos, que más o menos se ajustan a una recta:" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "import numpy as np\n", "import matplotlib.pyplot as plt\n", "%matplotlib inline\n", "import seaborn as sns; sns.set()\n", "\n", "rng = np.random.RandomState(42)\n", "x = 10 * rng.rand(50)\n", "y = 2 * x - 1 + rng.randn(50)\n", "plt.scatter(x, y);" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**Paso 1**: Importamos la clase adecuada para modelar los datos." ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "from sklearn.linear_model import LinearRegression" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**Paso 2**: Escogemos los hiperparámetros del modelo. Dependiendo del que hayamos elegido tendremos que responder a determinadas preguntas. En este caso vamos a fijar el parámetro `fit_intercept` a True para que la línea trazada por nuestro modelo no tenga que pasar necesariamente por el punto (0,0); es decir, la intersección con el eje *y* será determinada por la línea que mejor se ajuste a las muestras." ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None,\n", " normalize=False)" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model = LinearRegression(fit_intercept=True)\n", "model" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**Paso 3**: Acomodamos los datos en la matriz de características y el vector de etiquetas. En este caso el segundo ya lo tenemos (`y`), y para el primero simplemente tenemos que convertir el array unidimensional x en bidimensional." ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(50, 1)" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "X = x[:, np.newaxis]\n", "X.shape" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**Paso 4**: Aplicamos el modelo a los datos y obtenemos la pendiente (el coeficiente lineal) y el corte con el eje (el offset)." ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "model.coef_: [1.9776566]\n", "model.intercept_: -0.9033107255311164\n" ] } ], "source": [ "model.fit(X, y)\n", "print('model.coef_:', model.coef_)\n", "print('model.intercept_:', model.intercept_)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**Paso 5**: Predecimos las etiquetas para nuevas muestras, una vez que el modelo ya ha sido entrenado. En este caso tendremos unas observaciones x, y habrá que obtener la predicción para los valores de y." ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Generamos muestras de forma arbitraria\n", "xfit = np.linspace(-1, 11, 50)\n", "# Ajustamos igual que hicimos con x\n", "Xfit = xfit[:, np.newaxis]\n", "\n", "# Obtenemos las predicciones\n", "yfit = model.predict(Xfit)\n", "\n", "# Visualizamos el resultado (dibujar la línea que une las predicciones es dibujar la recta de regresión)\n", "plt.scatter(x, y)\n", "plt.plot(xfit, yfit);" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "NOTA: tendríamos que evaluar la eficacia de nuestro modelo (no lo hacemos con este ejemplo puesto que los datos no son reales).\n", "\n", "### Ejemplo de aprendizaje supervisado: Clasificación\n", "\n", "Usamos el dataset de iris importado previamente, usando parte de las muestras para entrenamiento y el resto para test. Esto se puede hacer de forma manual, pero usaremos el método `train_test_split()`. Para el modelo optaremos por aplicar el algoritmo Naive Bayes Gaussiano; muy rápido y sin hiperparámetros a escoger. Se usa para realizar una primera clasificación antes de explorar otras opciones.\n", "\n", "Aplicamos los mismos pasos que antes:" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [], "source": [ "from sklearn.model_selection import train_test_split #0\n", "from sklearn.naive_bayes import GaussianNB #1\n", "\n", "model = GaussianNB() #2\n", "X_train, X_test, y_train, y_test = train_test_split(X_iris, y_iris, random_state=1) #3\n", "model.fit(X_train, y_train) #4\n", "y_model = model.predict(X_test) #5" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Ahora sí podemos evaluar el error. Acudimos al método `accuracy_score()` para medir la precisión del modelo:" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.9736842105263158" ] }, "execution_count": 11, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.metrics import accuracy_score\n", "accuracy_score(y_test, y_model)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Ejemplo de aprendizaje no supervisado: Reducción de la dimensionalidad\n", "\n", "Aplicaremos reducción de la dimensionalidad al dataset de iris para poder visualizar mejor los datos. En nuestro caso aplicaremos PCA, que es una técnica lineal rápida, para pasar de 4 a 2 dimensiones. Seguiremos los mismos pasos que siempre:" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(150, 2)" ] }, "execution_count": 12, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.decomposition import PCA #1\n", "model = PCA(n_components=2) #2\n", "\n", "model.fit(X_iris) #4\n", "X_2D = model.transform(X_iris) #5 Transformar los datos a 2 dimensiones aplicando el modelo\n", "X_2D.shape" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Comparamos la visualización original con la nueva, y vemos que en la representación bidimensional las especies están bastante bien separadas, a pesar de que PCA no conoce las etiquetas. Esto indica que la clasificación será fácil." ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "sns.pairplot(iris, hue='species', height=1.5);" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [ { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAawAAAFcCAYAAACUWoq1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3Xl8lPW1+PHP8zyzJSELZAHUxgrKjaKp4gLKFVxYZZWLVEVUqrZal/5s1aJYF9RaAUWtS6m3t71S6xWRQnFBxYV7SykuqBEFUVoaEQnZl5nM9jzP749hhplkkkySWZPzfv36umaSmTkEfjn5fr/ne45imqaJEEIIkebUVAcghBBCxEISlhBCiIwgCUsIIURGkIQlhBAiI0jCEkIIkREkYQkhhMgIkrCEEEJkBElYQgghMoIkLCGEEBlBEpYQQoiMIAlLCCFERpCEJYQQIiNYUvXGLS0tXHzxxfzmN7/hqKOOivjcE088wUsvvUReXh4A8+bNY/78+TG/dm1tC4aRmJ6+AwdmU1/vSshr94TE0zmJp3PpFE86xQKxxVNcnJukaASkKGF98skn3Hnnnezduzfq53fs2MEjjzzCKaecktzAYmCxaKkOIYLE0zmJp3PpFE86xQLpF49I0Zbg6tWrufvuuykpKYn6+R07drBy5UpmzJjBkiVL8Hg8SY5QCCFEulFSOQ/rvPPO49lnn43YEnQ6nfy///f/WLRoEUcffTSLFi3iyCOP5Oabb05VmEIIIdJA2iWstj7//HPuuOMO1q1bF/PrJvIMq7g4l+rq5oS8dk9IPJ2TeDqXTvGkUywQWzxyhpVcaVcluH//ftasWRP62DRNLJaU1YYIIYRIE2mXsBwOB8uWLePrr7/GNE2ee+45Jk6cmOqwhBBCpFjaJKxrrrmGTz/9lEGDBrFkyRKuu+46pkyZgmmaLFy4MNXhCSGESLGU7rW9/fbbof9+5plnQv89efJkJk+enIqQhBBCpKm0WWEJIYQQnZGEJYQQIiNIwhJCCJERJGEJIYTICHLBSQiR9j6r2cWmys3UuusodAxiQul4RhaVpToskWSywhJCpLXPanaxevc6Gr1NZFuyaPQ2sXr3Oj6r2ZXq0ESSScISQqS1TZWb0VQNu2ZDURTsmg1N1dhUuTnVoYkkky3BFKvYU8PGbZXUNLopyncwZXQp5cOLUh2WEGmj1l1HtiUr4jGbaqXWXZeiiESqyAorhSr21PDcm7tpcHrJdlhocHp57s3dVOypSXVoQqSNQscgvIYv4jGv4aPQMShFEYlUkYSVQhu3VaJpKnarFtjqsGpomsrGbZWpDk2ItDGhdDy6oePRvZimiUf3ohs6E0rHpzo0kWSSsFKoptGNzRL5V2CzqNQ0ulMUkRDpZ2RRGfNGzCbflofL30q+LY95I2ZLlWA/JGdYKVSU76DB6cVuPTyK2+s3KMp3pDAqIdLPyKIySVBCVlipNGV0Kbpu4PHpga0On46uG0wZXZrq0IQQIu3ICiuFgtWAUiUohBBdk4SVYuXDiyRBCSFEDGRLUAghREaQhCWEECIjSMISQgiRESRhCSGEyAiSsIQQQmQESVhCCCEygiQsIYQQGUESlhBCiIwgF4cTSGZdCSFE/EjCSpDgrCtNUyNmXQGStIRIgM9qdrGpcjO17joKHYOYUDpeGub2MbIlmCAy60qI5PmsZherd6+j0dtEtiWLRm8Tq3ev47OaXakOTcSRJKwEkVlXQiTPpsrNaKqGXbMFfkHUbGiqxqbKzakOTcSRJKwEKcp34PUbEY/JrCshEqPWXYdNtUY8ZlOt1LrrUhSRSARJWAkis66ESJ5CxyC8hi/iMa/ho9AxKEURiUSQhJUg5cOLmD9xBAU5NlxuPwU5NuZPHCEFF0IkwITS8eiGjkf3Bn5B1L3ohs6E0vGpDk3EkVQJJpDMuhIiOUYWlTGP2VIl2MdJwhJC9Akji8okQfVxkrCSQC4QCyFE70nCSjC5QCyEEPEhRRcJJheIhRAiPiRhJZhcIBZCiPiQhJVgcoFYCCHiQxJWgskFYiGEiA8pukiwYGGFVAkKIUTvSMJKArlALIQQvSdbgkIIITJCyhJWS0sL06dPZ9++fe0+t3PnTubMmcPkyZNZvHgxfr8/BREKIYRIJylJWJ988gmXXHIJe/fujfr5W2+9lbvuuovXX38d0zRZvXp1cgMUQgiRdlKSsFavXs3dd99NSUlJu8998803uN1uTj75ZADmzJnDxo0bkx2iEEKINJOSoosHHnigw88dPHiQ4uLi0MfFxcVUVVV16/ULCwf0OLZYFBfnJvT1u0vi6ZzE07l0iiedYoH0i6e/S7sqQcMwUBQl9LFpmhEfx6K2tgXDMOMdGhD4B1xd3ZyQ1+4JiadzEk/n0imedIoFYotHElpypV2V4JAhQ6iurg59XFNTE3XrUAghRP+SdgnryCOPxG638+GHHwKwfv16xo0bl+KohBBCpFraJKxrrrmGTz/9FIDly5fz4IMPMmXKFFwuF5dffnmKoxNCCJFqKT3Devvtt0P//cwzz4T+u6ysjDVr1qQiJCGEEGkqbVZYQgghRGckYQkhhMgIkrCEEEJkBElYQgghMkLaXRwWh1XsqZE5WkIIcYgkrDRVsaeG597cjaapZDssNDi9PPfmbgBJWkKIfkm2BNPUxm2VaJqK3aqhKAp2q4amqWzcVpnq0IQQIiUkYaWpmkY3NkvkX4/NolLT6E5RREIIkVqSsNJUUb4Dr9+IeMzrNyjKd6QoIiGESC1JWGlqyuhSdN3A49MxTROPT0fXDaaMLk11aEIIkRJSdJFAFXtqWPPuHqrqXIDC4IEO5p57bExFE8GvkSpBIYQIkISVIBV7avivV3fhbPUCCopi8m1dK//1yk5+MO34mJOWJCghhAiQLcEE2bitErfHj6KoaKqCqigoCri9ulT6CSFED8gKK0FqGt3ohoGmHv6dQAF0w5RKPyGAz2p2salyM7XuOgodg/iP8ikcZTk61WGJNCYrrAQpynegqSqGefgxE9BURSr9RL/3Wc0uVu9eR6O3iWxLFo3eJn734Qt8VrMr1aGJNCYJK0GmjC7FYbdgmga6YWKYJqYJDpsmlX6i39tUuRlN1bBrtsDFeM2GVdXYVLk51aGJNCYJK0HKhxfxgwvKGFqYg6qAgsLQQVkxF1wI0ZfVuuuwqdaIx2yajVp3XYoiEplAzrASqCdVftLwVvQHhY5BNHqbsGu20GNe3UuhY1AKoxLpTlZYaSTY8LbB6Y1oeFuxpybVoQkRVxNKx6MbOh7dG7gYr3vxGToTSsenOjSRxiRhpRFpeCv6i5FFZcwbMZt8Wx4ufyv5tjyuOvX7jCwqS3VoIo3JlmAaqWl0k+2I/CuRhreirxpZVBaRoIqLc6mubk5hRCLdyQorjUjDWyGE6JissNLIlNGlPPfmbjwEVlZevyENb/sZX2UFvorXMJqrUXOLsZZPxVpanuqwkqbtZeIJpeNlm1CEyAorjZQPL2L+xBEU5Nhwuf0U5NiYP3GEVAn2E77KCjxbVmE4G8CWg+FswLNlFb7KilSHlhQffbuj3WXi1bvXyWViESIrrDQjDW/7L1/Fa6BaUKz2wANWO6Yv8Hh/WGX9ZdebocvEAHbNhgcvmyo3yypLALLCEiJtGM3VYLFFPmixBR7vBw46a9tfJlatcplYhMgKK8nkYrDoiJpbHNgODK6wAPxe1Nzi1AWVRCU5hdS01EdeJjZ8cplYhMgKK4nkYrDojLV8Khh+TJ8H0zQxfR4w/IHH+4GZZRPbXSbW5TKxCCMJK4nkYrDojLW0HPvYBag5BeB1ouYUYB+7oF+cXwGcMvTEdpeJ542YLedXIkS2BJNILgaLrlhLy/tNgoqm7WViIcLJCiuJ5GKwEEL0nKywEiy8yMJh03C1+gC5GCyEEN0lCSuBgkUWmqaS7bAEVleKgkUBl9svVYJCCNENsiWYQG2LLHTdwO3xU1XfKslKCCG6SRJWAtU0urFZAt9il9tHXbMHwzQxTFNK2oUQoptkSzCBivIdNDi92K0aTS4fCgAKVktg1eUhsAqTVZboK6R5rUgkSVgJFN593X+oOlBRTPJyAp0MpKRd9CWf1exi9e51aKoW0bx2Hom7SyUJsn+RLcEECu++rioKmqowKM9Blj3we4KUtIu+ZFPl5lDzWkVRsGs2NFVjU+XmhLxfMEFKd/f+Q1ZYCRbsvh6sGFRVBdM0paRd9Dm17jqyLVkRjyWyeW14ggTp7t4fSMJKguBdLLdXR9f9WDSFI4pypEpQ9CmFjkE0epuS1rw22QlSpF5KEtaGDRt4+umn8fv9XHHFFcyfPz/i80888QQvvfQSeXl5AMybN6/d12SK8LtYdqtKk9eH22vQcugCsRCxyIRJxBNKx7N69zo8eLGpVryGL6HNa5OdIEXqJT1hVVVVsWLFCtauXYvNZuPiiy9m9OjRHHvssaGv2bFjB4888ginnHJKssOLu+BdLF03qG/xogCKonCwvpXn3twNIKss0angJGJUS8QkYkivxrgji8qYx+ykFUEkO0GK1Et6wvrb3/7GmDFjKCgoAGDy5Mls3LiRG264IfQ1O3bsYOXKlXzzzTecfvrp/PznP8dut3f0kmkt2PC2qtmDAqiKAgoYphnq1C4JS3QmkyYRJ7N5bbITpEi9pCesgwcPUlx8eCBdSUkJFRUVoY+dTifHH388t956K0cffTSLFi3iqaee4uabb052qHERvIvl9xuogYtYGCZYNFXK2kVMjOZqsOVEPtiPJhF3Rrq79y9JT1iGYaAoSuhj0zQjPs7JyeGZZ54JffyDH/yAO+64o1sJq7BwQHyC7UBxcW7MX/v9SWWsXFuBRVMwDBPz0OMD8xwYpsnQ4gHder3expMMEk+Ac892Greux99wEEtBCflnzoLiUd2OxzdoCP6WelTL4SsQhs+NZdCQuPzZuvMaH327g7/sepODzlpKcgqZWTaRU4ae2OsYehJLMqRbPP1d0hPWkCFD+OCDD0IfV1dXU1JSEvp4//79/O1vf2Pu3LlAIKFZLN0Ls7a2BcMwu/7CHiguzqW6ujnmrz+6KJuLzz+WNe98xf5aFxZNoSDXFpio6jU4/5QjuvV6vY0n0SSegIhzJ0sWekMtB1/5LSXTfogr77juvdjxk9C3rEL3G2Cxgd8Lhh/L8ZN6/Wfrzvcn/GKwQ7VT01LPM+89H7chi739u4r3JeJY4pGEllxJvzh81llnsXXrVurq6mhtbeWNN95g3Lhxoc87HA6WLVvG119/jWmaPPfcc0ycODHZYcZV+fAillw9hpvmljPsiDxMEwpybMyfOELOr/qo8HMnRVEC50+qhcat67v9WukyiTjZF4O7Qy4R9w9JX2ENHjyYm2++mcsvvxyfz8fcuXMpLy/nmmuu4aabbuKkk05iyZIlXHfddfh8PkaNGsXChQuTHWZCBC8Rxyp8lpZ0d88sHZ07+RsOYu3B66XDJOJ0vvckl4j7h5Tcw5oxYwYzZsyIeCz83Gry5MlMnjw52WElXHcSUNtZWsHu7iBl8JlAzS3GcDaANay61e/FUlDS8ZPSXDrfe0rnZCriR3oJJkkwATU4vREJqKPxIm1nadmtWqgMXqQ/a/lUMPyYPg+maWL6PGD4A4UXGWpC6Xh0Q8ejewNnsLo3be49FToG4TUiL+OnSzIV8SOtmZIkPAEB2K0ajR4/v/3L52Q7LO1WXMH7W+GkDD5zBLbvFrTrTpEzfBSuKAf5yehkEa0o4Zzi02N+fjrfe5JLxP2DJKwkaZuAXG4fza0+ME2KChzttvzCZ2kFSXf3zBLruVMyOll0NPojvyCLoyxHx/w66XrvKZ2TqYgfSVhJ0jYBNbl8YILVcnjLr8mrh1ZcDquKy+0HAisr6e7edyWjk0VHRQl/2fUmPz7x6ri8R6qlazIV8SMJK0nChznaLCo+v4GCSV5O4AdIq8dPk9ODiUJRgQOv3wDTxKIquNx+qRLsw5LRyaKjooSDztpuv1Zv7jvJwEXRG5Kwkshu1aiqcwEKVk0h22ELDXNscnoDj1vU0IoLYECWlSVXjU5d0CLhOqooVHOLO35SN3VU4VcyoLBbSaQ3U4VTMZFY9C1SJZgEwQpBv2EytCiHwgIHdquGbph4fDqmaeLz66BAXvbhWzpSZNE/dFRRaC2fGrf36KjCb2TJiG5duO3N5eG2zzVMg0ZvM7/d8SyPbV/JR9/uiNufV/RNssJKgmgVggAWBQZk26hpdGO3WrBZVbIdhxOWFFn0Dx1VFMazSrCjooTNB/7arQu3Xd136my1Fv7cVr+bendDqLdmo7eJ3334AnOHz5TVluiQJKwk6KhE3eX2s+TSUcDhVZjHp0uRRT+UjE4W0YoSXtyzHocaObonmICiJZ/OLg93teUX/twmbwsoCopphhKmjl86U4hOScKKs2jdLGIpUQ8WU0grpr4hEyYEA5TkFFLTUt8uATk0R9TkM3rIaWw78EHU+05dtUcKvyvl032hKQ251sB0BZtmk84UolOSsOIodFalm7jcPuqb3ez5ponT/q2I2iZ3qEKwo9VTd3sNivQUvFdlGjp4nOgt9ehVX6GfPB2mXJbq8CLMLJvIM+893y4BaaoWNfl82bCHeSOi33d6YfefO90ujNiWbK1HURTybblkWwPP8epe6UwhOiVFF3G0cVslft2k2eVFN8AwwKcbbP38IPXNHnw+HZfbL53a+zhfxWuBZNXaBKYBqgamge/jl3Hu2Z7q8CKcMvRE5o2YTb4tD5e/lXxbHvNGzMZjeLCpkW16u+rNF0t7pJFFZfxk1I/44UmXk2/LRVO1UBGITzpTiC7ICiuOahrduNw+QEHXDcIncvl1k/oWDzPP+i4z/31YqkIUSWA0V4PHCYoS+B8AKhgGjVvXY518S0rjayva2VZhZfSzKrtq7/CcqjvtkaIVgfxH+ZRudd0Q/Y8krDgqyndQ3+xGU1X8UT6voPDG+/skYfVxam4xekt9YGUVYoJmwd9wEDLgfKuj5GPRLB2eU/1k1I+61R6pbaJMt+GfIv3IlmAcTRldiqqo6FGmHSsKqAq4vdFSmehLrOVTQVXB1ME0A9uCJmDLRrFl4dmyKnBROKxvoK+yItVhRxhZVBZ1q9Ctu6NuFX7rrOKx7St5YfefAfj+iAv5yagfScWfiCtZYcVBxZ4a1rzzFVX1bnTDIEq+QlMVDBMcNvmW93XW0nL0k6fj+/jlwEGmZgF7NopmAYW49A1MRhVirFuFzT4nbt3d7vKxdLAQ8SY/PXupYk8N//XKTpxu/6EjCwUFk0P/DwBNPfSLNiaTTj8qpfGK5HCcOgut+Jh2ScW/9Y/QppKuu30Dk9HdvSPRtgpbfE5yLNkZNe1XehpmJklYvbRxWyVur35oy08BBRRTQVUgL8eGy+3H7fXjsGlMOv0oOb/qR6JeBt5Zgt5Q22XfwM5WUG27u+/K0ths06nb/SeKa7Ym9IdvtGIJp89Fnm1AxNel87Rf6WmYuSRh9VJNoxvdMFGVw4+pCuiGAcATN49LUWQiHeWfOYuDr/wW0wdYbOD3tusb2NUKKry7+y6Ln3VZPjQTsnQ9KT98224VPrZ9ZYfdL9JRVxecRfqShNVLRfkOmpxeDNMkmLMMEzRVDXWyiNb9Qu5g9T++ygoad76B6feA7gRVQxt0VLvzp67mY4V3d99s96OZYDNBUS1YUvDDN9Om/XbVD1GkL0lYvTRldGnoDMvAxDQBTBwOG1NGl4a6X2iaSrbD0m6ysOgfgqsmzWZDyR4YsbJqu23Y1Xwsa/nUQCcNH9SpJlmGCZgo2flA73749uRsJ9Om/XbWD1GkNylr76Xy4UX8YNrxDB2UhULg7GpoYQ4/uKCM8uFFEZ3ag3OuNE1l47bKVIcukii4alKtjkBhjtUOqiXweBtqbnEgoYULO+eylpZjH7sANaeAQX4Dv6qiDihEsR1qcdTDH77Bs51YR42EC3awWHLW7Wlfzt7RqJV0XRGKw2SFFQed9QDsqFO7zLnqezorlOjOVOHwFVRH51zBgo7Jh5KMV9WwmWavtuP6y9lOpq0IxWGSsBKss07tcrbVd3RVKHH43Cns/8t1MFW4O/OxuvvDt+2WX3g7pI7OdvY7D/DY9pV96od7tDtmIv1JwkqwKaNLA3OuiOzUXlZaIGdbfUhXhRLBVZPhc2NiibpqCted+Vix/vCNVs4dPjQx2tlOk7cFt+6RS8EiLcgZVhLYLSq1DW6+rXFiURXmTxzBrsoGOdvqQ4zm6sD2XbjwQolD506WAQPB60TNKcA+NvEXfcNFG29vDRtvH+1sx+V3McCaE/EcLew5IrNdc801uN2ZczzR5Qpr7dq17N+/nwkTJlBWdvg3qpUrV/KjH/0oocFluvAKwaFF2Xj9Bh6fDsjZVl8TXmoe0mbLz1paTvGpY1PW4DXqll/Y0MRo24suXyu51sizNykB7zueeeaZVIfQLZ0mrGXLlvHJJ59w3HHHcfnll/OLX/yCGTNmALBx40ZJWF0IrxAEsFs1PIcej2UKsUhv4UUWWB2YXlfgEx0USqRa1HLuNkMTM/1ScF/2zTffcMstt9DS0sIRRxzBRRddxAsvvIDL5aK6upqFCxdyySWX0NLSwuLFi/n666/JyspiyZIlDB8+nHfeeYdHHnkEXdeZPn06P/7xjznvvPPYsGEDpmlGfc59993HBx98gGEY3HHHHZx55pkp/R50mrA2b97M2rVrsdlsXHTRRVx11VUUFxczZswYTDNKh9cMlogCiM5WUZdNGhH1bKvtFGKRntoWWeD3gmmiaBZMrzMtx4ZEu+BrKiYTjglUFEa7g5Vpl4L7srVr1zJnzhwuuugi/vCHP/DFF1/wySef8Morr+BwOLjwwgs599xz+e///m+mTZvGpEmT+Oyzz7j77rv53e9+x7333ssLL7xAYWEhCxcuZM6cOaHXfvLJJ9s951e/+hWff/4569ev58svv+TVV19N74Slqio2W+A3qxNOOIGlS5fys5/9jBdeeAFFUTp7akZJ1OXejlZRDqsa6kGo634smsIRRTlSJZhBohVZACiOAeTMvS+FkXWss6GJHfbXGzGbeSOkBDwdnHHGGdx8881s376dKVOm4HA4+OqrryguDmw7n3XWWXz00Uds27aNv/71rzz55JMAtLS0sHv3boYPH87gwYMBWLVqVcRrR3vO4MGDaWlp4YorrmDSpEn84Ac/SOKfNrpOE9bRRx/NY489xmWXXUZhYSFnn302V199NQsXLsTr9Xb21IzS2dZdbxJItApBV6sPFAW/CQNz7RErK0lWmaM796rSwav/2MTb+/4Xj9+L3WLjvKPGccGwCaGhiZ3dwUr3i8D9xejRo1m3bh2bNm3i0Ucf5cILL8RiOfwj3DRNNE1D13VWrlzJEUccAUBVVRV1dZFnjgcOHCAvLy/0cbTnWK1WXnrpJbZs2cLrr7/OmjVr+POf/5yEP2nHOq0SvPfee9m3bx/bt28PPbZw4UKuvPJKmpv7zmTQmkY3Nkvkt6I3BRAVe2pY+qft/PGN3ditGhYFXG4/BTk28nJsZDssUh2Y4brqRpFOXv3HJl7buwmv7kNVVLy6j9f2buLVf2wKfU2tuy7qYEYprkgfS5cuZevWrVx66aUsXLiQP/7xj/z973+nqamJpqYmtm3bxqhRoxg1ahQvvvgiAG+//TY333wzxxxzDHv37qW6uhq/389NN93E/v37Q68d7Tk7d+7khhtu4JxzzuGee+6hqqoKXddT8mcP6nSFNWjQIJYtW9bu8fnz5zN06NCEBZVs8SyAaLu9GFxBXTZpBOXDi7jt6b91eK4lF4kzRyzdKNLF2/v+F0VR0JTgL2UKOgZv7/tfrhh9ISD99TLBJZdcwi233MIf/vAHBgwYwLXXXsuzzz7LD3/4QxobG7n++uspKiripptu4o477mDGjBnY7XZ+9atf4XA4uOuuu7j66qvRdZ0LL7yQY489NvTa0Z5z7LHHctRRRzF9+nRsNhu33XYbmqZ1EmHiKWY3qifcbjd//vOfefbZZ6mpqeH9999PZGw9VlvbghFt7G8HwpNMeAHE/Ikj2iWM4BZKR5b+aXu75Ofx6RTk2Ljt0lEdft6iKnh8ekwxdCeeZOtP8fRk6m8qvj83vr0IVVED89oOMUwTwzT4n+8/SXV1c8QZVnhxxbwRybsgnIn/doqLc5MUTXvbtm3jueee4/HHH09ZDMkWU6eLqqoq/vjHP7J69WqcTic//OEPufLKKxMcWvIEE0I8Vjdd3a/qqPOFRVETco4mEqc73SjCJXq8fdtqP4tqQTd14HDCMjGxh110lv56IhN0mrAqKir4wx/+wFtvvcWZZ57J3XffzfLly7npppuSFV/SdNbAtiPRtvC62l7sKDn+8Y3dcpG4H3Du2Z7Q8fbRqv00RcVn+ABQUDAxMU2T846KHC4q/fUyy+jRoxk9enSqw0iqThPWvHnzmDlzJps2bQqVTj788MNJCSzddVQKP/bEIWzZcaDT+1XRkmNRfqVcJO4HGreu77TnYG9Fq/bLs+eSpWfRqre2qxIUIpN0WiW4ePFiPv/8c+bMmcOyZcv48ssvkxVX2utoztWuygbmTxxBQY4tVBnY1TkUBLYKdT3Qusk0TTw+XS4S90H+hoOd9hzsrY6q/XyGj+8MOJIcazYAm7/ZwmPbV/LRtzvi8r5CJEOnK6wFCxawYMECtm3bxvPPP8+FF16IqqqsW7eO6dOnR9wB6G86O6vqyfZiPM/RRPqyFJSgN9R22nOwN6JV+zX7nLh1Nwdba3D5XKAogI8qV3VEt3Yh0l1MGSe4V1pTU8MLL7zAY489xqOPPsq7776b4PDSVyJ6AfYk0YnMkn/mLA6+8tuElcNHa6XU4nOSY8mm1e9GUVVUFAzTwK27GWDP6nMDGkXf1a0lUlFREddffz3XXXcd77zzTqJiyggdVfu13cKTu1V9T2+q/HKGj8I+NrbhjD0RrdrP6XORZxtAi9OJeuguloKC39AjurWL/mX16tVkZ2czffr0VIcSsy4T1uOPP87pp58eanp42223UVpayg033NDjN92wYQOBCddlAAAgAElEQVRPP/00fr+fK664gvnz50d8fufOnSxevBin08lpp53Gvffem3bbj7Fs4SWqR6FIna4mC8eip+Xwseqo43qwvD1YKWhRtXbd2kX/sX37ds4444xUh9EtnWaBZ555hi1btjB79uzQY5dddhkPPPAAOTk5LFy4sNtvWFVVxYoVK0Jd4C+++GJGjx4dcev61ltv5f777+fkk0/mjjvuYPXq1Vx66aXdfq9E62oLL1E9CkXqdDVZOB0FtwmzLA6aPc3ohy4QOzQHPkMPdWsXyfPBzirWvvsVVXUuBg/KZs45x3La8YN7/boHDhzglltuweVyoaoqd955J6qq8uCDD+J2uxk4cCD33nsvX3/9NW+//TZ///vfKS4u5vjjj2fx4sXs378fi8XCzTffzLhx49i6dWuo21F+fj4PP/wwgwYNYsWKFWzdupXGxkZKSkpYsWIFRUWJ/5nWaZXghg0b+O1vf0tpaVhJdnk5Tz75JGvXru3RG/7tb39jzJgxFBQUkJ2dzeTJk9m4cWPo89988w1ut5uTTz4ZgDlz5kR8PpPEu0ehSL2uJguno5FFZcwbMZuSrCKyrdnYNCtZFgeDs4u56tTv9/r86rOaXTy2fSV3/e1BHtu+ks9qdsUp8r7pg51VrFxbQX1TK7lZFuqbWlm5toIPdlb1+rXXrFnDOeecw9q1a7npppt4//33ufPOO3n44Yf585//zMKFC/nFL37BWWedxXnnncdNN93E2WefzX333ceYMWPYsGEDjz/+OHfccQc1NTU89dRT3HPPPaxdu5azzjqLzz//nH/961/84x//4H/+5394/fXXGTp0KH/5y1/i8J3pWqcrLE3TyM/Pb/d4UVFRj3tKHTx4MHSnC6CkpISKiooOP19cXExVVe//IlNBhjT2PbFMFk5HHV0K7m07pA7HkpC8lk6ZZu27X2GxKDhsgR+/DpsFN37WvvtVr1dZZ555JjfeeCM7d+5k/PjxjB8/nqeeeorrrrsu9DUtLS3tnvf3v/+d+++/H4DvfOc7fO973+OTTz7h/PPP54YbbmDChAmcf/75jB07FoCf//znvPjii/zzn//k448/jljUJFKXB0Nerzc0Eyv8Mb/f36M3NAwjYpaWaZoRH3f1+VgUFg7oUWyxirV/2PcnlbFybQW6YQS2A306mIHHw1+jt9sDqexnFk1fjsc5bg61G/8TTB+KxY7p94BiUDhuDjlh7+Pcs53GrevxNxzEUlBC/pmzyBk+Ku7xxENv4tm846/YrVbslkACt2LB4/ew+cBfOef405MaSyIkIp6qOhe5WZE/eu1WjYN1rl6/9qmnnsorr7zCu+++y6uvvsqLL77IUUcdxfr164HAGJGampp2z2vbUtY0TXRd58orr+Tcc8/lnXfeYdmyZVRUVHD22Wfzs5/9jCuvvJLJkyejqmrSBvp2mrDOOeccli1bxh133BFKGqZpsnz5csaMGdOjNxwyZAgffPBB6OPq6mpKSkoiPl9dfXh7paamJuLzsehu89vu6M5vpEcXZXPx+ce2K8w4uig79BrhhRkOq0p1vYun1nwc02Xj7saTDH0+nrzjsIyZj6/iNfSwKj9X3nG4Dr1PRGGGJQu9oZaDr/wW+9gFHHHq2Iz8/kSbRjyyqIwDTdVkW7Lw+w+PnVBNjQNN1d3+c2biv52eJLTBg7Kpb2oNrbAg0AC7ZFB2t1+rraVLlzJ48GCuuOIKRo8ezaxZs8jKyuKDDz7gtNNO46WXXmLDhg2sWrUqNDsLYMyYMaxZs4aFCxfy9ddfs337du655x4uuugi7r33Xq688koKCgp46623cDgcnHHGGVxyySXU19fz7rvvMmnSpF7HHotOE9Z1113Hj3/8Y84//3y+973vYRgGn376KcOGDePXv/51j97wrLPO4te//jV1dXVkZWXxxhtvcN99hye0Hnnkkdjtdj788ENOPfVU1q9fz7hx4zp5xfQmhRl9T1dVfp0VZnDq2CRFGT+dbfvJWJLum3POsaxcW4Ebf2jnxe83mXPOsV0/uQsLFizgZz/7GWvXrkXTNJYtW0Z+fj4PPPAAHo+HAQMG8NBDDwGBn8WPPPIIubm5LF68mLvuuitUm3D//fdTUlLCT3/6UxYtWoTFYiE7O5v777+frKwsbrjhBmbMmAHAiSeeyL59+3odeyy6HC+ya9cuNm3ahKIo5OTkcPLJJ4cKInpqw4YNrFy5Ep/Px9y5c7nmmmu45ppruOmmmzjppJPYtWsXd955Jy0tLYwcOZIHH3yw3bZkZ9JlhRWL4HysttugLrefpdedlfR4eiuT44lXF/WW528BW067v1O8To65aWXGfX+CZfHhScmje8m35YUqEOMxliQT/+30dMsweAxwsM5FSRyrBPu6TldYL730Eg899BBHH300lZWVLF++vNfJCmDGjBmh7Bz0zDPPhP67rKyMNWvW9Pp9MoEUZqSHeNyvCsrUwoyO1LrryLZkRTwWnEYsY0l65rTjB0uC6oFOE9aqVavYsGEDgwcP5qOPPmLFihWcffbZyYotY/Smm0WsHTNEYsXzflUs04gTPRMrnrra9pOxJCJZOr2HBTB4cOC3gFNOOYX6+vqEB5RpgkUTDU5vRDeLij3tK3GiKR9e1KPu7iK+4nm/ylpajn3sAtScAvA6UXMKsI89vFILruYMZ0PEas5XWdHFK6fGhNLx6IaOR/cGJgnoXnRDZ0KpXDgWydXpCqttOXlP7171ZfEompCmt6nX0TaeYs3C9fJD3V4JdVaYkWndMmTbT6SLbjXo6+59qP6gozEj39Q4Wfqn7dL0NkNE28Yzva7APUDdH9fpwEZzNdhyIh88tJrrqHw81WTbT6SDThPWF198wahRo0Ifu91uRo0aFbrMu3379oQHmO6iFU00Ob24PToH6ly0enTqmj3s2b+DaWNKmfnvw1IYrehIIAFFdlE3NA10vdcrofDzKt+gISjWLEy/t91q7ov8PP4iXSOE6FCnCevNN99MVhwZK1rRhNPtx2ZVaW71oQCaAoZh8srWSr47NE9WWmmq7TZesDw9QjfPtdpWH/pb6jFczYeGKBJRlPG/A3PQFCNivL0Hr8yrEuKQThPWkUcemaw4Mk54ZaDDqoKi4HL7Kcp34Gr14fEZKIB66AeTpoBuGKx5d4/Mx8oQ8ShPb3tepVocKHYDNA3VkRtxNla/7xWytejl4+m6VSj6tueffx6ASy65pFvPW7t2Le+99x6/+tWv4hpPeg2ZyhBt51x5/Qa6X+eySYHqvqV/2s7ufY1oYUd+hhk4A9xf46RkULbMx8oAsZSnd6Wj8yq8TrLn3h/xcGHN1qjl43bVLg1mRUp0N1ElmiSsHuiqMnDK6FL27N+BYZiB7UATwMQELNKGKWNEO9fqrEow2t2qrlZp4Ssnh+bA5WsFiOgaYdEsaKomW4V9SGfNkXsj2DJp8uTJQGA80z333MOKFStoaGjA4XDwi1/8ghNOOIFFixbR0NDAv/71L2699Vbef/99tmzZgqqqTJgwgRtuuCHUgu/GG28MDd5VFIWTTjqJ++67D7/fz5133skXX3yBoihcddVVEfMTAT7++ONQa6iBAweyZMkSjj76aBYsWEB+fj5ffvkljz76KMcff3yXfz5JWD3QUWVgTaM7tFWoqeD3mximic2iku2w0djiIS/PFvV5Ij3FOh24o04ZlhH/jrH7r6FVmuFzh1ZpbXv0NfucuPytOH0uVEVlcHYx/zFiBi/s/nOHnSZ6IjxJDskrZvyQf5fEl0TOPdsDHf81C4pjAP6W+sDHU67uddKaNWsWGzZsYPLkyezduxePx8Mvf/lL7rrrLk444QS++uorrr/+el5//XUACgoK+M1vfsM333zDww8/zCuvvEJrayu33347Ho8n9LpVVVU8+OCDrF27liFDhnDrrbeyefNmtm/fzsCBA3n55Zepq6vjoosuoqzs8L8lr9fLT3/6Ux599FHKy8t57bXX+OlPf8pLL70EwL/927/xxBNPxPzn6/LisGivKN+B129EPOb1GzhsWugSccnAbPIH2EIrqsEDsziiMBuLRWv3PGnDlPnCz6oURUGx2jENHd+nr2P6PZitjZjOOiwDBoYuEW+q3BxaObl1Dy3ewJwiq2ahMGsgHj3wA6PQMQiv4Yt4v542mA0myUZvE9mWLBpaG1m9e50MXUyixq3rQbOgWh0oioJqdYBmCTzeS+PHj+ejjz6ipaWFl19+mQsuuIAdO3Zw++23M2vWLH72s5/hcrlCTSDKywO/jA0ePBi73c7FF1/Ms88+yy233ILdfnhX4KOPPmLUqFEMGTIEgGXLljFhwgT+/ve/M3fuXAAGDRrE+eefz3vvvRd63t69e8nLywu9z9SpU6msrKS5uTni/WMlCasHpowuRdcNPD49cPPfp6PrBphmaKtQURSsFhUFaPUEZoedVlYS9XnShinzte2UYXpbwdUIPg9K9kCUrHwUq4P8M2eFVmy17jpsqhWAJm8LKAoqCn5Dx67Z0FSNTZWb49ppIjxJKoqC3WIPvY9IDn/DQRSLPeIxxWLH33Cw169ts9k499xzefvtt9m4cSOzZ8/GZrOxfv360P9efPFFCgoKAHA4Ar8sWywWXnzxRX7yk5/Q0NDAxRdfzD//+c/Q61oskQ266+rqqKur63COVpBhRP5i3/Zrgu8fK0lYPdBROyW3z8BmCXxLXW4fdc0eDNPEr5v8Y38Tf9myF9MEi4K0Yepj1NziQFHGIWZrIyiAxRpacaFG/hYdvnLyG34UFExMLGpgFW5TrXzrrGJT5WbcuocmbzMNnibybXk96oYOkUkyyKZa2e88IGPuk8RSUBIY/BnG9HuwFHRv7l9HZs2axe9//3sKCgo48sgj+e53vxsa4Lhlyxbmz5/f7jmff/45l112Gaeffjo///nPGT58eETCOumkk/j4449Dswp/+ctf8tZbb4XmaEEgib311lucccYZoecNGzaMhoaG0FT5V199lSOOOCKUMLtLzrB6KFo7paL8ytAl4iZX4A6WYQSKLYxDl62bnF6UAbZQRaFID71tRtu+otAXSFiOvMNfZLHhbzhIMF0ER3N48KIpGroZ+K0z1xqYmN3sc+LW3TR6mxhozw8VYcRa0h6tFD5aI9smbwtu3UOVqxq37qbe08g/d/yLSaXnccGwCTF/D0Rs8s+cRe3G/8TAfXhqte4n/8xZcXn9U089lebm5lCF37Jly7jnnnv4z//8T6xWKytWrGjXteiEE07g5JNPZvr06WRlZTFq1CjGjRvHZ599BgS2DBcvXsxVV12FYRicfPLJzJkzh9bWVu655x5mzJiBrutce+21jBw5ki+++AIIrPhWrFjBfffdR2trK/n5+axYsaLHf7Yu52FlomTOw4q4j2XTaHJ6yc6yUl0fqPYyDBNVBU0NrLwM06SoIIuCHBu3Xdr7qqBMnCGUTLHEEzkh+HD5enjD2liEJz3T4wKLHTU7P/R50+fBXlCIdfItoceCSeVbZxVu3U22JZs82wC8ho9adz05lmzy7YdnLgXnUP1k1I86jSW8oCO84nD0kNPYduCD0OOGonPQWYtVteHRPSiAgoJB4BesH554edIKMjLx305P52Elqkqwr5MVVi+E38dSFDhY34pPN3B7A78pa6qCCaiHfpkxzEBZu1QGppd4NaMNrygMJkHT54lIgvlnzsIV9pzwHn1tV0ROn4s824CI94i1OjD8rAoOl8J/2bCHeSNmR1QJNrmdeA3voYvugV+sVBR0Q5fS+QTJGT5KElQPSMLqheB9LMMwqW/2AAqqAj790EGjwaFTCQXz0F2svBy7VAammc6a0fZUR3e4coaPwtXBb+1tG8xGm/Qba3VgV0MXg+9TXJzLna8v56vGf6KFHWkHztIs0mVDpBVJWL0QvI91sL6VwIGFSTBXaapy6CEFv25isyjkD7CjqopUBsZRPAYhJmpCcKx3uDoSfsYVvq0XS3VgV0MX277PP3f8C8M0UQ/9ioVpkm3LwqE5pMuGSBtSJdgLwftYft1AVUA/dG6mKIGEpQDFA7M4qjiHYUfkYZpIZWAc9XYQoq+yAtfLD6HX7cNsqcFobcI0zcA2XjdbMCXCyKIy5o2YTb4tD5e/tVvVgd0phR9ZVMak0vNQlMA2oIrKANsALIqGaZqRZfBh5fZCJJussHoh2KldVRR0wyRYvqIqkedVLrefJVeNTm2wfVBvzp7CCy2UnEGYrU3gbsY0/GgDj0ybkfU9nUPV3aGLFwybwNF5R7X7+nh32RCiNyRh9UJwlbTm3T3sr3GiHNoFVBQVOa9KvN6cPbVNdkp2PqbVgZpTQPb0nyci3B7pzflRd5NdtK8vrIx9a1GIRJMtwV4qH17EkqtGc9Pcco4szAYCK6yBuXJelWhtL+sCMZ89te1MAWDqfvSqr2h5/hZcLz8U89ZiorRtoxQ8P0rmpd54dtkQqbF48WI+/fTTmL/+rbfe4rHHHovra8aL3MPqpq7uZoTfy0rGvKtMvLsSL7Hcn+ooHtfLD2E4G0IrLMPjgpZaUDWUgqE9vovVle58f6JVCcZ6Dyue8SSrSjAT/y339B6W6BnZEoyjZCer/q434z+wOjC9h25EWWzgagBAyRkY6ALQw7tY8dRZaXoy9fQcTXTso2938Jddb3LQWUtJTiEzyyZyytATe/260caL/Otf/+Kpp54CAl0vDMPguOOO48477+S2226jsrKS73znOxw4cIAnnniC9957LzR88bzzzmPmzJn89a9/pbW1lYceeogTTzyRBQsWcMMNN3DGGWewfPlyNm3ahKZpfP/73+eKK67gvffeY8WKFbjdbpqamrj99tuZMKH3XVMkYcVJ26GOMpwxOXo6/gO/F0wTRbNgep2ACQMKUWxhCaKXd7F6IjypFhTZaHHo2B2Hf4uX86PM99G3O/jdhy9gVTUGWLNpaG3kdx++wFWn0uukFW28yAknnBD6/N69e3nnnXfIzc3lV7/6FccccwxPP/00n376Kd///vejvmZBQQFr1qxh1apVrFy5MjQjC2Djxo1s376dDRs24PP5uPTSS7ngggv44x//yP3338/w4cPZunUrv/zlL+OSsOQMK07ChzoqioLdqqFpKhu3VaY6NEH78R+YBvjcGI0HUHOLUfOHoKiRo1/icRerWzG2KdMf16zja23E426W86M+5C+73sSqatgt9lDHfKuq8Zddb/b6tduOF5k5c2bE54855hhycwO/AG3ZsoVZswL9C0866SRGjBgR9TXPPvtsAI477jgaGhoiPvf+++8zdepUbDYbOTk5rF+/nuLiYpYtW8aXX37Jk08+ye9//3ucTmev/2wgCStmFXtqWPqn7Vz1wJss/dN2KvbURHy+ptEd6tQeJC2Y0kd4kYXpbcVsqQ0kLdNAbziA0bAfs+kARv1+DLczJXex2ibV48liVovKALer2/ewRPo66KzFprUZ5KrZOOis7fVrtx0vMn369IjPh4/z0DSt3XiQaIJzsdo2zIX2Y0f27duHy+Xi0ksvpaKighNPPJFrr722p3+c9u8Xt1fqw8K3+3Kzom/3FeU7Qp3ag6SkPX2Ed7MIjP4IdCbBVMF96GBdOfR356yFgUdgP/OSmM7Detpho61oZfplpoOyGicDJt7eq9cW6aMkp5CG1kbsYTOxvLqXkpzCuLz+rFmzuP/++0PjRTpy5plnsmHDBsrKyvjiiy/48ssvoyalzpx++uk8++yzXHzxxfj9fq6++mqWLl3K3r17+dOf/oTNZmP58uURM7J6Q1ZYMYhlu6+joY5S0h4/wc4UPSk7t5ZPBcMfWDn5fWCaYHLo4hyhZKUOPAIlbzCqI7fTZNWbDhsd6U2ZvsgcM8sm4jN0PH5P4GeF34PP0JlZNjEurx8cL9J2O7Ct66+/nsrKSmbMmMHjjz9OUVFRtwcqTpw4kVGjRjFnzhzmzp3L5ZdfTnl5OXPnzmXatGlMnToVp9OJ2+3G5XJ1/YJdkLL2GNz29N/IdlhCU4R9fgPTNHG5/Sy97qzQ16WiSjATS4F7oqcjQMLjCa6K9KqvAAWyCwKrqUMdylE11PwhgW0Sr5MBlyyP+pptS+IhMDoETUN15Ha66urs+xPrnzGeZebp9O8nnWKBxJa1J6pKsDvWr1/PUUcdxamnnsr+/fu57LLL2LRpE6qavusY2RKMQazbfdGGOor4iMcIkGBF4eHEoGGqFjD8gIKSdWh2VRermmhbd6buh6YqjLzBEasuiP0eVyxl+uFzrqQZbeY6ZeiJSU9QbQ0bNoy7774bwzBQVZUlS5akdbICSVgxCfYM9AAWTZHtvhSI5wiQ8MSge1rAZ4AjN3A3K4Zii6jd3Vsb4z5TK5qO5lzJ3CrRXSeddBJr165NdRjdkt7pNE2UDy9i/sQRFOTYaGn1S8f1FEjU+Y5itaPkD0a154DXiZpT0OU2Y/h5WHh3d7ILIr8wAfe4at112FRrxGPSjFb0F7LCilFwuy/d9tn7C2v51EMTfIk43+lJ2Xm0S8Sm4cc+9vLYV0NWO2ZjFUDgDtfAIwPbguF6mVCjVSJ2Z86VEH2NrLBERrCWlmMfuwA1pyDmlVBH2t53Uqx2UC2Bx7t6bjDZ6TpKwREoA4owfW60Y06Luurq6T2ujioRz8kqlWa0ot+SFZbIGL2d4BsUz7EkwbMq49td2MfG3tewp+9z3J4K5o2Jfc6VEH2JJCzR70QtmujOWJIOkl28EmpX7yPNaEV/JVuCot/pqGgilu27ZF3ulUvEQrQnCUv0O705D+tNsutWjEl6HyEyiWwJin6pp9t33Z3B1Zv4kvE+QmQSSVjdULGnhrfWVPBtdYsMaOyH2paZd6sMvgfieSYmRF+Q9IS1f/9+br31VmpraznmmGNYvnw5OTmRh8vffPMN06dPp7Q00EmiqKiI3/3ud8kONUKwY7vdpsmAxn6o7d2tnrReSqZkjbUXIpmSnrDuvfdeLr30UqZNm8aTTz7JU089xa233hrxNTt27GDGjBksWbIkaXF11bg22LHdYbPg8xvYrRqeQ49LwuobOhsZEo9ehski/QZFX5XUogufz8f777/P5MmTAZgzZw4bN25s93Wffvopu3fvZtasWVx++eV88cUXCY0ruHpqcHojVk/hQxplQGPf1tXIkPABkCEJaL0UD+H9BhVFwa7Z0FSNTZWbUx2aEL2S1BVWfX09AwYMwGIJvG1xcTFVVVXtvs5utzNz5kwuvvhi/u///o/rr7+eV199FZvN1u5royksHNCtuN5aU4HdpuGwBeKyWTXcXj9vfbSf88ccA8DQ4gHUN7UCYD2UuNxeP0OLB/R4xEC8pPr928rEePa//gaazYZqPdSB32rB8Llh5xsUnzoW36Ah+FvqUS2HO/QbPjeWQUO6/edN9Pen3tfAAFt2xDA+TXNQ72uI+t7p9PeVTrFA+sXT3yUsYb322ms8+OCDEY8dffTR7SZaRptweeONN4b+e/z48Tz88MP84x//oKwstu2M7s7D+ra6hWxHYKsvSFUUvq1uCfUNPP+UI0JnVqqi4PUb6LrB+acckdLegunW2zBT4/HUHQisrPyHJ6OaWNDrDgSef/wk9C2r0P1GRC9Dy/GTuvXnjef3p6NzqoHWgnb9Bj26l4G2gnbvnU5/X+kUCyR2HpbomYQlrKlTpzJ1auSdEZ/Px+jRo9F1HU3TqK6upqSkpN1zV61axfTp0xk4cCAApmmGVmWJEMu8q+A51Vsf7ZcqwT4i/MzK9LjAMFCy8g5/QdhF3Z6Wmbc9F3OOmwN5x/U69s7OqSaUjmf17nV48GJTrXgNn/QbFH1CUrcErVYrp512Gq+++iozZsxg3bp1jBs3rt3Xvf/++7jdbq655hree+89DMNg2LBhCYsrfN6VzaKGVk9t512VDy/i/DHHpNVvgaJn2nVs13VwNWAAiiM3ajf47paZR6ssrN34n1jGzO91oUZnc7F+MupHzEP6DYq+J+lVgnfffTeLFi3i6aefZujQoTzyyCMAPP/88xw8eJCf/OQnLF68mEWLFrF+/XrsdjsPP/xwQidhBldJyR5vL1KnbdWfkp2PAeBzg6rG5aJutMpCTF9cKgtr3XVkW7IiHgufiyX9BkVflPSEdeSRR7Jq1ap2j19yySWh/x48eDC///3vkxlWt8fbd1UGL9JbtOaySlYeeJ0MuGR54t7DYkfvZmXhq//YxNv7/heP34vdYuO8o8bJXCzRL0mnix4IlsFrmiqXiNOYr7KC/a+/gafuQLsVU286tscq2nuYfk+33uPVf2zitb2bUBQFVVHx6j5e27uJU0u+R527Xs6pRL8izW97IHiJ2G7VAvdcrBqaprJxW2WqQxOHBM+P/C31Ue9VJaO5bNT30Lv3Hm/v+18URUFTVNRD/1dRFHbU7WTeiNnk2/Jw+VvJt+Uxb4RcDBZ9m6yweqCm0U22I/JbJ5eI00vw/Ei1OgKl6m06UySjuWy09ygcNwdXmyrBztooefxeVCXy90oFBY/fK+dUot+RhNUDsZTBi9SKZapwMprLtn2PnOJcXGFVpl21UbJbbHh1H3D4vqKJib1t1w0h+gHZEuyBKaNL0XUDj0/HNE08Pj1qGbxInUwZgNhVG6XzjhqHaZropoFx6P+apsl5R7W/DiJEXycJqwfKhxcxf+IICnJsuNx+CnJszJ84Qgou0kjw/MjwudN6AGKtuw6bao14LLw8/YJhE5j63QnYNCuGaWDTrEz97gQuGDYhFeEKkVKyJdhD3S2DF8kVPD9i5xvoUaoE00Us5ekXDJMEJQRIwhJ9mLW0nOJTx6Z1ZxJpoyRE7CRhCRFFZ7Ox4mlkUZm0URIiRpKwRMbqSVKJ5TnJni4s5elCxEaKLkRG6mrgYm+eE94DUFGUQC9A1RJ4XAiRMpKwREbqSVKJ9TmZNF1YiLyTizMAAA7nSURBVP5EtgRFRop2Mdg0dPSqr2h5/pbQdh/FYzt9TrRElIw+g0KI7pMVlshIbS8Gm95WaKkFFExU9Ko9uF9/lK9/+9PQll+sl4mT0WdQCNF9krBERmqXVJz1gU/YssBZB5igqPjrvw2dU8WaiKyl5djHLkDNKQCvEzWnAPvYxBRcCCFiJ1uCIiO1bSwLJuQMAk8zKErgf4BpGCiHzqmyp/+cWBveJqPPoBCieyRhiYwVnlRcLz+E4WzA1P0Q6m5uolgsEedUkoiEyFySsETcJevSbThr+dTAXSlFBdMIPGiCml2AkWEFE52NGxGiP5MzLBFXzj3bu30/Kh5C5075gw8lLBWyB6JoWkYVTATHjTR6myLGjXxWsyvVoQmRcrLCEnHVuHV96K4T0G5wYiIFt/vCV3iWAQOxHD8pY7YBg+NGDNPgYGstfsOPqqis3/OarLJEvycJS8SVv+EgWLIiH0zypdvwc6ri4ty0bn7bVq27DgWFBk8jKAqqomKYBt86q/isZpckLdGvyZagiCtLQUlGDE5MV4FxI4FKRxUFhcCs4fChjkL0V5KwRFzlnzlLLt32woTS8ehGYJK1aZoYpoEJ5FtzQ0MdheivJGGJuMoZPkou3fbCyKIyhmSXBLYCMdFUjYH2fDRNixjqKER/JGdYIu7S+a5TKkruu2v2sRewevc6NFWToY5ChJEVlug3ejKSJBVGFpUxb8Rs8m15uPyt5NvymDdithRciH5PVlii3wgfLwIkteS+u2SooxDtyQpL9Bsy50qIzCYJS/QbsY4XEUKkJ0lYot+QOVdCZDZJWKLfkDlXQmQ2KboQ/Uo6l9wLITonKywhhBAZQRKWEEKIjCAJSwghREaQM6xeqthTw8ZtldQ0uinKdzBldCnlw4tSHZYQQvQ5krB6oWJPDc+9uRtNU8l2WGhwennuzd0AkrSEECLOZEuwFzZuq0TTVOxWDUVRsFs1NE1l47bKVIcmhBB9jiSsXqhpdGOzRH4LbRaVmkZ3iiISQoi+SxJWLxTlO/D6jYjHvH6DonxHiiISQoi+SxJWL0wZXYquG3h8gQmxHp+OrhtMGV2a6tCEEKLPSVnRxaOPPoqmadx4443tPuf1elm8eDE7duzA4XCwfPlyhg8fnoIoOxcsrJAqQSGESLykJ6zm5mYefPBBXnnlFa6++uqoX7Nq1SqysrJ47bXXeP/997n99ttZvXp1kiONTfnwIklQMQpO+9Xr9oGhg2ZBG3hkWk79FUKkn6RvCb711lt897vfZeHChR1+zbvvvsvMmTMBOP3006mrq2P//v3JClEkQHDar95wALwu8HvA7URvrErLqb9CiPST9BXW7NmzAfj1r3/d4dccPHiQ4uLDM4qKi4s5cOAARxxxREzvUVg4oHdBdqG4ODehr99dmRDP/tffQLPZ0JubMRUVRVExTQPF50LLLYSdb1B86tikxZNKEk/H0ikWSL94+ruEJazXXnuNBx98MOKxYcOG8Yc//KHL55qmiaIoER+rauyLwdraFgzDjPnru6O4OJfq6uaEvHZPZEo8nroDYMvB9PtAUTExAQXT70fHgl53ICF/jkz5/qRKOsWTTrFAbPFIQkuuhCWsqVOnMnVqzwbjDR48mIMHD1JaGqi2q6mpoaSkJJ7hiSRTc4sxnA2gWcHwAwpggmaRqb9CiJikZVn7+PHjWb9+PQAffPABdrs95u1AkZ6C036xZoFpBoouDANs2TL1VwgRk7RJWM8//zyPPfYYAAsWLMDr9TJt2jQeeOABli5dmuLoRG8Fp/1qBUMCScpiB0cOWv5gmforhIiJYppmYg57UkjOsFJH4umcxNOxdIoF5AwrHaXNCksIIYTojCQsIYQQGUESlhBCiIwgCUsIIURGkIQlhBAiI0jCEkIIkREkYQkhhMgIkrCEEEJkBElYQgghMoIkLCGEEBkh6fOwhMg0n9XsYlPlZmrddRQ6BjGhdDwji8pSHZYQ/Y6ssIToxGc1u1i9ex2N3iayLVk0eptYvXsdn9XsSnVoQvQ7krCE6MSmys1oqoZds6EoCnbNhqZqbKrcnOrQhOh3JGEJ0Yladx021RrxmE21UuuuS1FEQvRfkrCE6EShYxBewxfxmNfwUegYlKKIhOi/JGEJ0YkJpePRDR2P7sU0TTy6F93QmVA6PtWhCdHvSMISohMji8qYN2I2+bY8XP5W8m15zBsxW6oEhUgBKWsXacVXWYGv4jWM5mrU3GKs5VOxlpanNKaRRWWSoIRIA7LCEmnDV1mBZ8sqDGcD2HIwnA14tqzCV1mR6tCEEGlAEpZIG76K10C1oFjtKIqCYrWDagk8LoTo9yRhibRhNFeDxRb5oMUWeFwI0e9JwhJpQ80tBr838kG/N/C4EKLfk4Ql0oa1fCoYfkyfB9M0MX0eMPyBx4UQ/Z4kLJE2rKXl2McuQM0pAK8TNacA+9gFKa8SFEKkBylrF2nFWlouCUoIEZWssIQQQmQESVhCCCEygiQsIYQQGUESlhBCiIwgCUsIIURGkIQlhBAiI0jCEkIIkREkYQkhhMgIkrCEEEJkBElYQgghMkKfbM2kqkpGv353STydk3g6l07xpFMskH7x9HeKaZpmqoMQQgghuiJbgkIIITKCJCwhhBAZQRKWEEKIjCAJSwghREaQhCWEECIjSMISQgiRESRhCSGEyAiSsIQQQmQESVhCCCEygiSsHvjggw+YM2cOM2bM4Nprr6WxsTFlsXz44YfMnTuXWbNmccUVV/DNN9+kLJZwjz76KL/+9a9T9v4bNmzgggsuYNKkSTz33HMpiyNcS0sL06dPZ9++fakOhSeeeIJp06Yxbdo0li5dmupweOyxx7jggguYNm0av//971MdDgAPPfQQixYtSnUYIpwpum3ChAnml19+aZqmaS5btsx8+OGHUxbLueeea+7cudM0TdN88cUXzWuvvTZlsZimaTY1NZm33367WV5ebj7++OMpieHAgQPmueeea9bX15tOp9OcMWNG6O8rVT7++GNz+vTp/7+9uwtpsgvgAP6PuVleiF7oCgJXGSbF6DutcBSxENtHEWQEsrpombAaZB8SlpBZLVBECJIIP4jSIoTQEGd10WRQFBLSCCtKGmv2IdRsz9ye90Lc+65PilfPZv8f7OLMjf2fq//Oedw58uLFi+XXr18LzXL//n15x44dcjAYlCVJkktKSuTu7m5hedxut1xcXCyHQiF5dHRU3rBhgzw4OCgsjyzLssvlktesWSMfOXJEaA6KxRnWH+js7ER2djZCoRB8Ph9SU1OF5JAkCQcOHMCiRYsAADk5OfB6vUKyTHA6ndBoNNi9e7ewDC6XC3l5eUhLS0NKSgo2b96M27dvC8sDAG1tbThx4gQyMzOF5gCAjIwMHD16FCqVCkqlEgsWLMCbN2+E5Vm9ejWam5uRlJSEd+/eIRwOIyUlRViejx8/ora2Fvv27ROWgb6PhfUHlEolPB4PdDod3G43ioqKhORQqVQwmUwAgEgkgoaGBmzatElIlglmsxl79+6FQqEQluHt27fIyMiIjjMzM+Hz+YTlAYDq6mqsXLlSaIYJCxcuxNKlSwEAL1++RFdXF3Q6ndBMSqUS9fX1KCoqQn5+PtRqtbAslZWVsNvtwr6I0o+xsH6iq6sLBQUFMQ+LxQJgfDbjcrmwf/9+2O12oVkkScKhQ4cwNjYGq9U66Vl+lUe0SCSCGTP+PRZCluWYMY179uwZ9uzZg8OHD0Oj0YiOA5vNhr6+Pni9XrS1tQnJ0N7ejjlz5iA/P1/I59PPTcvzsP4vhYWFKCwsjHkuGAyip6cnOpMxGo04e/askCwA8PnzZ5SWliItLQ0XLlyAUqmc9Cw/yxMPZs+ejQcPHkTHfr8/Lpbi4snDhw9hs9lQUVEhbIVgwuDgICRJQm5uLmbNmgW9Xg+PxyMkS2dnJ/x+P0wmE0ZGRhAIBHD69GlUVFQIyUOxOMP6TUlJSaiqqsKTJ08AjM80li9fLixPeXk5srKyUFdXB5VKJSxHPFm7di36+vrw/v17jI6Ooru7GwUFBaJjxQ2v14uysjKcP39eeFkBwNDQEI4fPw5JkiBJEpxOJ1asWCEky+XLl3Hr1i10dHTAZrNh48aNLKs4whnWb1IoFKitrUVlZSXC4TDUajWqq6uFZBkYGIDT6UR2dja2bt0KYPx+TWNjo5A88UKtVsNut6OkpAShUAjbt2+HVqsVHStuXLp0CcFgEGfOnIk+V1xcjJ07dwrJo9Pp0N/fD7PZDIVCAb1eHxdFSvGHJw4TEVFC4JIgERElBBYWERElBBYWERElBBYWERElBBYWERElBBYWTRtDQ0PIzc2FyWSKPoxGI65fvw5g/EffdXV1MJvNMJlMMBgMuHjxIr7+R9mWlhbk5OTg8ePH3/2cp0+fYv369ZN+PUQUi7/Domll5syZ6OjoiI59Ph+2bNmCJUuWwOFwYN68ebh27RqSk5Px4cMHWK1WBAIBHDx4MPqeq1evwmAwoKmpKbrnHgCMjY2htbUVjY2NCAQCU3pdRMQZFk1zarUaWVlZ6OnpwfPnz3Hs2DEkJycDANLT03Hu3DmsWrUq+nq3242RkRGUl5fD6XTG7H4/MDAAj8eDhoaGKb8OImJh0TT36NEjvHr1CgCg1Wq/2UVeo9Fg3bp10fGVK1dgMBigVquRl5eH1tbW6N+0Wi1qampidoInoqnDJUGaVr58+RI9ciUcDiM9PR0OhwMvXrz45l7V1/x+P5xOJ27cuAFg/KiUkydPoqysTOj5TEQ0joVF08rX97AmpKamoqmpCeFwOGaW1d/fj5aWFjgcjuiRFqWlpQDGjyn59OkTbt68iV27dk3NBRDRD3FJkP4Ky5Ytw/z581FTU4NgMAgAGB4exqlTpzB37lyEw2G0t7ejqqoKvb296O3txd27d2G1WtHc3PzL2RkRTT4WFv016uvrAQDbtm2D0WiExWKBXq+HzWbDnTt3EIlEYDAYYt5jsVgwPDyMe/fuiYhMRP/B3dqJiCghcIZFREQJgYVFREQJgYVFREQJgYVFREQJgYVFREQJgYVFREQJgYVFREQJgYVFREQJ4R8p4KMKJu86+gAAAABJRU5ErkJggg==\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Insertamos los resultados en el DataFrame original\n", "iris['PCA1'] = X_2D[:, 0]\n", "iris['PCA2'] = X_2D[:, 1]\n", "# Visualizamos las 2 dimensiones nuevas\n", "sns.lmplot(\"PCA1\", \"PCA2\", hue='species', data=iris, fit_reg=False);" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Ejemplo de aprendizaje no supervisado: Agrupamiento\n", "\n", "Aplicaremos clustering al dataset de iris para intentar agrupar los datos sin usar las etiquetas. En nuestro caso usaremos GMM (Gaussian Mixture Model), que intenta modelar los datos como una colección de manchas gaussianas." ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from sklearn.mixture import GaussianMixture #1\n", "model = GaussianMixture(n_components=3, covariance_type='full') #2\n", "\n", "model.fit(X_iris) #4\n", "y_gmm = model.predict(X_iris) #5 \n", "\n", "iris['cluster'] = y_gmm\n", "sns.lmplot(\"PCA1\", \"PCA2\", data=iris, hue='species', col='cluster', fit_reg=False);" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Ejemplo: Reconocimiento de dígitos manuscritos\n", "\n", "Escogemos ahora un problema más interesante de OCR (Optical Character Recognition) para clasificar números escritos a mano. Partiremos del dataset disponible en Scikit-Learn, con datos ya preformateados (1797 muestras de 8x8 píxeles)" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(1797, 8, 8)" ] }, "execution_count": 16, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.datasets import load_digits\n", "digits = load_digits()\n", "digits.images.shape" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "import matplotlib.pyplot as plt\n", "\n", "fig, axes = plt.subplots(10, 10, figsize=(8, 8),\n", " subplot_kw={'xticks':[], 'yticks':[]},\n", " gridspec_kw=dict(hspace=0.1, wspace=0.1))\n", "\n", "for i, ax in enumerate(axes.flat):\n", " ax.imshow(digits.images[i], cmap='binary', interpolation='nearest')\n", " ax.text(0.05, 0.05, str(digits.target[i]),\n", " transform=ax.transAxes, color='green')" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Trataremos cada uno de los 64 píxeles asociados a cada caracter como un atributo distinto. Tenemos disponibles la matriz de características en digits.data, y el array de etiquetas en digits.target.\n", "\n", "#### Reducción de la dimensionalidad\n", "\n", "Para intentar visualizar los datos en 2 dimensiones (en lugar de usar 64), usaremos **Isomap** (algoritmo de Manifold Learning):" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(1797, 2)" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.manifold import Isomap #1\n", "iso = Isomap(n_components=2) #2\n", "X = digits.data; y = digits.target #3\n", "iso.fit(digits.data) #4\n", "data_projected = iso.transform(digits.data) #5\n", "data_projected.shape" ] }, { "cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "plt.scatter(data_projected[:, 0], data_projected[:, 1], c=digits.target, \n", " edgecolor='none', alpha=0.5, cmap=plt.cm.get_cmap('tab10', 10))\n", "plt.colorbar(label='digit label', ticks=range(10))\n", "plt.clim(-0.5, 9.5);" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Podemos observar por ejemplo lo bien separados que están el 0 y el 1, y por el contrario la confusión entre el 2 y el 7, o el 3 y el 9. Pero más o menos se puede comprobar que están razonablemente separados como para poder usar un algoritmo de aprendizaje supervisado.\n", "\n", "#### Clasificación\n", "\n", "Usaremos de nuevo Naive Bayes Gaussiano:" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.8333333333333334" ] }, "execution_count": 20, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.naive_bayes import GaussianNB\n", "model = GaussianNB()\n", "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)\n", "model.fit(X_train, y_train)\n", "y_model = model.predict(X_test)\n", "\n", "accuracy_score(y_test, y_model)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Obtenemos más de un 80% de efectividad, que no está nada mal. Para mejorar es interesante saber dónde se equivoca nuestro modelo; algo que podemos ver con la matriz de confusión, o volviendo a pintar el grid de caracteres con el valor estimado en rojo si no coincide con el real. Para mejorar los resultados deberíamos optar por otros algoritmos!" ] }, { "cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from sklearn.metrics import confusion_matrix\n", "\n", "mat = confusion_matrix(y_test, y_model)\n", "\n", "sns.heatmap(mat, square=True, annot=True, cbar=False, cmap=\"YlGnBu\")\n", "plt.xlabel('predicted value')\n", "plt.ylabel('true value');" ] }, { "cell_type": "code", "execution_count": 22, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "fig, axes = plt.subplots(10, 10, figsize=(8, 8), subplot_kw={'xticks':[], 'yticks':[]},\n", " gridspec_kw=dict(hspace=0.1, wspace=0.1))\n", "\n", "test_images = X_test.reshape(-1, 8, 8)\n", "\n", "for i, ax in enumerate(axes.flat):\n", " ax.imshow(test_images[i], cmap='binary', interpolation='nearest')\n", " ax.text(0.05, 0.05, str(y_model[i]), transform=ax.transAxes,\n", " color='green' if (y_test[i] == y_model[i]) else 'red')" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Validación del modelo\n", "\n", "La **precisión** del modelo **no sirve** para validarlo correctamente. Podemos obtener un 100% de precisión con un modelo si por ejemplo lo probamos con los mismos datos que usamos para entrenarlo (¡menuda idea!).\n", "\n", "### Conjuntos de retención\n", "\n", "Una forma más correcta de validar el modelo sería usando un conjunto de retención, que no es más que un subconjunto de los datos que se reserva para la fase de validación del modelo (datos de test o validación). La desventaja de usar este procedimiento es que perderemos una parte de nuestros datos que nos vendría muy bien para entrenamiento (sobre todo cuando no tenemos muchos). Para evitarlo podemos recurrir a la validación cruzada.\n", "\n", "### Validación cruzada\n", "\n", "Con este método de validación, podemos dividir de varias maneras los datos disponibles en sendas iteraciones, de tal forma que creemos diferentes combinaciones de subconjuntos para entrenamiento y validación.\n", "\n", "![CV](images/py3_skl_cv.png)\n", "\n", "Por suerte con Scikit-Learn podemos hacer la validación cruzada directamente usando el método `cross_val_score()`. Eso nos dará una mejor idea del rendimiento de un algoritmo." ] }, { "cell_type": "code", "execution_count": 23, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "array([0.94871795, 0.94871795, 0.91666667, 1. ])" ] }, "execution_count": 23, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.model_selection import cross_val_score\n", "from sklearn.datasets import load_iris\n", "iris = load_iris()\n", "X = iris.data\n", "y = iris.target\n", "cross_val_score(model, X, y, cv=4)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Scikit-Learn implementa [varios esquemas de validación cruzada](https://scikit-learn.org/stable/modules/cross_validation.html) que son útiles para distintos casos. Todos ellos están disponibles en el módulo `model_selection`. Por ejemplo podríamos querer crear el máximo de iteraciones posibles:" ] }, { "cell_type": "code", "execution_count": 24, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.9533333333333334" ] }, "execution_count": 24, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.model_selection import KFold\n", "scores = cross_val_score(model, X, y, cv=KFold(len(X)))\n", "scores.mean()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Seleccionando el mejor modelo\n", "\n", "Este es uno de los aspectos más importantes en la práctica. Si nuestro estimador no rinde bien... ¿cómo deberíamos proceder? Existen varias posibles respuestas, aunque lo complicado es saber cuál será la adecuada en cada caso:\n", " * Usar un modelo más complicado o más flexible\n", " * Usar un modelo menos complicado o menos flexible xD\n", " * Conseguir más muestras de entrenamiento\n", " * Conseguir más datos de otras fuentes para que las muestras tengan más variables\n", " \n", "#### El balance sesgo - varianza\n", "\n", "Encontrar el mejor modelo es básicamente encontrar el punto dulce en el balance entre sesgo (*bias*) y varianza. \n", "\n", "![Balance](images/py3_skl_bal.png)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Un modelo con **alto sesgo** se **sub-ajustará** a los datos por su simplicidad y poca flexibilidad, siendo incapaz de describir o representar la información contenida en las características de los datos. El error de entrenamiento y el de validación serán similares.\n", "\n", "Un modelo con **alta varianza** se estará **sobre-ajustando** a los datos por su complejidad, por lo que será muy sensible a errores o ruido, y no generalizará bien para nuevas muestras de entrada. El error de validación en este caso será mucho más alto que el de entrenamiento.\n", "\n", "#### Curvas de validación\n", "\n", "En esta gráfica vemos lo que se conoce como curva de validación:\n", "![VAL](images/py3_skl_val.png)\n", "\n", "En algún punto intermedio, la curva para el score de validación tendrá un máximo. Ese punto sería nuestro punto dulce; el balance óptimo entre sesgo y varianza.\n", "\n", "Vamos a aplicar validación cruzada con el fin de hallar la curva de validación para un modelo de regresión polinómica, que es una generalización del modelo lineal, donde la complejidad viene marcada por el grado, que está parametrizado (¡si cambia el grado no hay que cambiar el modelo!). En Scikit-Learn podemos implementar este modelo combinando la regresión polinómica lineal simple con un *preprocesador polinómico*, usando una tubería:" ] }, { "cell_type": "code", "execution_count": 25, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from sklearn.preprocessing import PolynomialFeatures\n", "from sklearn.linear_model import LinearRegression\n", "from sklearn.pipeline import make_pipeline\n", "\n", "def PolynomialRegression(degree=2, **kwargs):\n", " return make_pipeline(PolynomialFeatures(degree), LinearRegression(**kwargs))\n", "\n", "def make_data(N, err=1.0, rseed=1):\n", " # randomly sample the data\n", " rng = np.random.RandomState(rseed)\n", " X = rng.rand(N, 1) ** 2\n", " y = 10 - 1. / (X.ravel() + 0.1)\n", " if err > 0:\n", " y += err * rng.randn(N)\n", " return X, y\n", "\n", "X, y = make_data(40)\n", "\n", "X_test = np.linspace(-0.1, 1.1, 500)[:, None]\n", "\n", "plt.scatter(X.ravel(), y, color='black')\n", "axis = plt.axis()\n", "for degree in [1, 3, 5]:\n", " y_test = PolynomialRegression(degree).fit(X, y).predict(X_test)\n", " plt.plot(X_test.ravel(), y_test, label='degree={0}'.format(degree))\n", "plt.xlim(-0.1, 1.0)\n", "plt.ylim(-2, 12)\n", "plt.legend(loc='best');" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "¿Y para qué grado del polinomio tendremos el mejor balance sesgo - varianza? Podemos averiguarlo usando directamente el método `validation_curve()` de Scikit-Learn para obtener la curva de validación:" ] }, { "cell_type": "code", "execution_count": 26, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from sklearn.model_selection import validation_curve\n", "degree = np.arange(0, 21)\n", "train_score, val_score = validation_curve(PolynomialRegression(), X, y, 'polynomialfeatures__degree', degree, cv=7)\n", "\n", "plt.plot(degree, np.median(train_score, 1), color='blue', label='training score')\n", "plt.plot(degree, np.median(val_score, 1), color='red', label='validation score')\n", "plt.legend(loc='best')\n", "plt.ylim(0, 1)\n", "plt.xlabel('degree')\n", "plt.ylabel('score');" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Vemos claramente como el punto dulce para un balance óptimo entre sesgo y varianza se consigue usando un polinomio de grado 3.\n", "\n", "#### Curvas de aprendizaje\n", "\n", "El modelo óptimo dependerá generalmente del tamaño del dataset; en principio un conjunto de muestras mayor soporta un modelo más complejo. **La curva de validación depende por tanto de la complejidad del modelo y del tamaño del dataset**.\n", "\n", "A veces es interesante ver cómo se comporta el modelo cuando incrementamos el número de muestras del conjunto de entrenamiento, algo que podemos hacer progresivamente.\n", "\n", "En esta gráfica vemos lo que se conoce como curva de aprendizaje:\n", "![LEA](images/py3_skl_lea.png)\n", "La característica más notable de esta curva es la convergencia hacia una puntuación según crece el número de muestras. Llegará un momento en que añadir más muestras no servirá de nada si no aumentamos la complejidad.\n", "\n", "Scikit-Learn proporciona el método `learning_curve()` para obtener estas curvas de forma sencilla:" ] }, { "cell_type": "code", "execution_count": 27, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from sklearn.model_selection import learning_curve\n", "\n", "fig, ax = plt.subplots(1, 2, figsize=(16, 6))\n", "fig.subplots_adjust(left=0.0625, right=0.95, wspace=0.1)\n", "\n", "for i, degree in enumerate([2, 9]):\n", " N, train_lc, val_lc = learning_curve(PolynomialRegression(degree), X, y, cv=7, train_sizes=np.linspace(0.3, 1, 25))\n", "\n", " ax[i].plot(N, np.mean(train_lc, 1), color='blue', label='training score')\n", " ax[i].plot(N, np.mean(val_lc, 1), color='red', label='validation score')\n", " ax[i].hlines(np.mean([train_lc[-1], val_lc[-1]]), N[0], N[-1], color='gray', linestyle='dashed')\n", "\n", " ax[i].set_ylim(0, 1)\n", " ax[i].set_xlim(N[0], N[-1])\n", " ax[i].set_xlabel('training size')\n", " ax[i].set_ylabel('score')\n", " ax[i].set_title('degree = {0}'.format(degree), size=14)\n", " ax[i].legend(loc='best')" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Vemos cómo con un modelo polinómico de grado 2 las curvas pronto convergen, por lo que aumentar el grado sería la única forma de mejorar el rendimiento; algo que vemos que se consigue (aunque también necesitamos más muestras).\n", "\n", "Pintar la curva de aprendizaje nos puede servir en un determinado momento para tomar decisiones sobre cómo mejorar nuestro modelo." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Validación en la práctica: Cuadrícula de búsqueda\n", "\n", "En la vida real los modelos no son tan sencillos como pintar una línea, por lo que estas visualizaciones son mucho más complejas, y lo que buscaremos es simplemente el modelo que maximiza la precisión en la validación.\n", "\n", "Scikit-Learn proporciona la clase `GridSearchCV` para realizar este tipo de búsqueda de forma automatizada. Aplicamos esto al ejemplo de regresión polinómica, realizando la optimización para un grid de 3 dimensiones resultado de variar 3 parámetros del modelo:" ] }, { "cell_type": "code", "execution_count": 28, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "{'linearregression__fit_intercept': False,\n", " 'linearregression__normalize': True,\n", " 'polynomialfeatures__degree': 4}" ] }, "execution_count": 28, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.model_selection import GridSearchCV\n", "\n", "param_grid = {'polynomialfeatures__degree': np.arange(21),\n", " 'linearregression__fit_intercept': [True, False],\n", " 'linearregression__normalize': [True, False]}\n", "\n", "grid = GridSearchCV(PolynomialRegression(), param_grid, cv=7, iid=False)\n", "\n", "grid.fit(X, y)\n", "grid.best_params_" ] }, { "cell_type": "code", "execution_count": 29, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Obtenemos el resultado\n", "model = grid.best_estimator_\n", "\n", "plt.scatter(X.ravel(), y)\n", "lim = plt.axis()\n", "y_test = model.fit(X, y).predict(X_test)\n", "plt.plot(X_test.ravel(), y_test);\n", "plt.axis(lim);" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Scikit-Learn [permite más cosas](https://scikit-learn.org/stable/modules/grid_search.html), como por ejemplo paralelizar los cálculos, hacer búsquedas aleatorias, o especificar una función de scoring propia.\n", "\n", "## Ingeniería de características (Feature Engineering)\n", "\n", "En los ejemplos vistos con anterioridad los datasets contenían datos numéricos, limpios y formateados. En el mundo real difícilmente encontraremos esto. Uno de los pasos más importantes para aplicar Machine Learning es la Ingeniería de características, que consiste en tomar cualquier información de entrada y convertirla en numérica para poder construir la matriz de características.\n", "\n", "### Atributos categóricos\n", "\n", "Un ejemplo habitual sería el Sexo de una persona; 'Hombre' o 'Mujer'.\n", "Una forma de lidiar con este tipo de atributos es crear columnas extra en el dataset que indiquen si está presente o no la categoría concreta, usando como valor 0 o 1 (una columna para 'Hombre' y otra para 'Mujer' en nuestro ejemplo).\n", "Scikit-Learn proporciona la utilidad `DictVectorizer` para hacer esto de forma automática. Vemos un ejemplo:" ] }, { "cell_type": "code", "execution_count": 30, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "array([[19, 0, 1, 0, 80, 1, 0],\n", " [71, 0, 0, 1, 77, 0, 1],\n", " [22, 1, 0, 0, 63, 0, 1],\n", " [53, 1, 0, 0, 58, 0, 1]], dtype=int32)" ] }, "execution_count": 30, "metadata": {}, "output_type": "execute_result" } ], "source": [ "data = [\n", " {'edad': 19, 'peso': 80, 'sexo': 'Hombre', 'estado': 'S'},\n", " {'edad': 71, 'peso': 77, 'sexo': 'Mujer', 'estado': 'V'},\n", " {'edad': 22, 'peso': 63, 'sexo': 'Mujer', 'estado': 'C'},\n", " {'edad': 53, 'peso': 58, 'sexo': 'Mujer', 'estado': 'C'}\n", "]\n", "\n", "from sklearn.feature_extraction import DictVectorizer\n", "vec = DictVectorizer(sparse=False, dtype=int)\n", "vec.fit_transform(data)" ] }, { "cell_type": "code", "execution_count": 31, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "['edad',\n", " 'estado=C',\n", " 'estado=S',\n", " 'estado=V',\n", " 'peso',\n", " 'sexo=Hombre',\n", " 'sexo=Mujer']" ] }, "execution_count": 31, "metadata": {}, "output_type": "execute_result" } ], "source": [ "vec.get_feature_names()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "El problema de este enfoque es si el número de categorías de un atributo es grande (por ejemplo, la provincia). Pero como la mayoría de valores será igual a 0, podemos crear una salida comprimida usando el parámetro `sparse=True`. Muchos modelos aceptan este tipo de datasets de entrada.\n", "\n", "### Atributos textuales\n", "\n", "A veces tenemos la necesidad de convertir texto en un conjunto de valores numéricos representativos. Por ejemplo para textos donde queremos analizar el sentimiento, nos gustaría saber qué palabras aparecen.\n", "Una forma de tratar estos campos sería vectorizar el texto usando el número de apariciones de cada palabra, y crear una matriz comprimida donde cada palabra nueva es una columna extra. Para ello tenemos la utilidad `CountVectorizer`:" ] }, { "cell_type": "code", "execution_count": 32, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "<3x13 sparse matrix of type ''\n", "\twith 16 stored elements in Compressed Sparse Row format>" ] }, "execution_count": 32, "metadata": {}, "output_type": "execute_result" } ], "source": [ "texto = ['no me ha gustado nada de nada; muy mala',\n", " 'muy buena, una obra maestra',\n", " 'ni buena ni mala']\n", "\n", "from sklearn.feature_extraction.text import CountVectorizer\n", "\n", "vec = CountVectorizer()\n", "X = vec.fit_transform(texto)\n", "X" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Este enfoque tiene alguna problemática, y es que las palabras más usadas aparecerían más veces, y quizá en nuestro caso no es lo deseado. Para ello podríamos usar `TFidVectorizer`\n", "\n", "### Imágenes\n", "\n", "Una forma de codificar las imágenes sería usando el mismo enfoque visto para los dígitos manuscritos; usando simplemente los valores de los píxeles, pero dependiendo de la aplicación esto no será lo óptimo.\n", "\n", "### Atributos derivados\n", "\n", "Un tipo interesante de atributo es aquél que deriva matemáticamente de otros. Podríamos por ejemplo convertir un dataset con dos atributos x e y en una **regresión polinómica** simplemente convirtiendo la columna x en varias (x^2, x^3, ...)\n", "\n", "### Valores no disponibles\n", "\n", "Debemos reemplazar dichos valores con lo más adecuado para nuestro caso. Scikit-Learn proporciona el módulo `impute` con varias utilidades. Por ejemplo la clase `SimpleImputer` para un enfoque típico usando la media, la mediana o el valor más frecuente de la columna:" ] }, { "cell_type": "code", "execution_count": 33, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "array([[4.5, 0. , 3. ],\n", " [3. , 7. , 9. ],\n", " [3. , 5. , 2. ],\n", " [4. , 5. , 6. ],\n", " [8. , 8. , 1. ]])" ] }, "execution_count": 33, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from numpy import nan\n", "X = np.array([[ nan, 0, 3 ],\n", " [ 3, 7, 9 ],\n", " [ 3, 5, 2 ],\n", " [ 4, nan, 6 ],\n", " [ 8, 8, 1 ]])\n", "\n", "from sklearn.impute import SimpleImputer\n", "imp = SimpleImputer(strategy='mean')\n", "X2 = imp.fit_transform(X)\n", "X2" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Tuberías\n", "\n", "Para encadenar transformaciones podemos usar el objeto `Pipeline`, que aplicará los pasos especificados a los datos de entrada." ] }, { "cell_type": "code", "execution_count": 34, "metadata": {}, "outputs": [], "source": [ "# Ejemplo\n", "from sklearn.pipeline import make_pipeline\n", "\n", "model = make_pipeline(SimpleImputer(strategy='mean'),\n", " PolynomialFeatures(degree=2),\n", " LinearRegression())" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.2" }, "nbTranslate": { "displayLangs": [ "en", "es" ], "hotkey": "alt-t", "langInMainMenu": false, "sourceLang": "es", "targetLang": "en", "useGoogleTranslate": true }, "toc": { "base_numbering": "0", "nav_menu": {}, "number_sections": true, "sideBar": true, "skip_h1_title": false, "title_cell": "Mis notas: Machine Learning con Scikit-Learn", "title_sidebar": "Contents", "toc_cell": true, "toc_position": { "height": "calc(100% - 180px)", "left": "10px", "top": "150px", "width": "418px" }, "toc_section_display": true, "toc_window_display": true } }, "nbformat": 4, "nbformat_minor": 2 }