In [0]:
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
Objetivos de aprendizaje:
Al igual que en el ejercicio anterior, trabajamos con el conjunto de datos de viviendas en California, pero esta vez lo convertiremos en un problema de clasificación binaria al predecir si una manzana es de costo elevado. Por el momento, también volveremos a los atributos predeterminados.
El objetivo de nuestro conjunto de datos es median_house_value, que es un atributo numérico (de valor continuo). Podemos crear una etiqueta booleana al aplicar un umbral a este valor continuo.
Dados los atributos que describen una manzana, queremos predecir si se trata de una manzana de costo elevado. Para preparar los objetivos para entrenar y evaluar los datos, definimos un umbral de clasificación del percentil 75° para el valor mediano de las casas (un valor de aproximadamente 265,000). Todos los valores de las casas por encima del umbral se etiquetan como 1 y los demás, como 0.
In [0]:
from __future__ import print_function
import math
from IPython import display
from matplotlib import cm
from matplotlib import gridspec
from matplotlib import pyplot as plt
import numpy as np
import pandas as pd
from sklearn import metrics
import tensorflow as tf
from tensorflow.python.data import Dataset
tf.logging.set_verbosity(tf.logging.ERROR)
pd.options.display.max_rows = 10
pd.options.display.float_format = '{:.1f}'.format
california_housing_dataframe = pd.read_csv("https://download.mlcc.google.com/mledu-datasets/california_housing_train.csv", sep=",")
california_housing_dataframe = california_housing_dataframe.reindex(
np.random.permutation(california_housing_dataframe.index))
Observa que el código es levemente distinto del de los ejercicios anteriores. En lugar de usar median_house_value como objetivo, creamos un nuevo objetivo binario, median_house_value_is_high.
In [0]:
def preprocess_features(california_housing_dataframe):
"""Prepares input features from California housing data set.
Args:
california_housing_dataframe: A Pandas DataFrame expected to contain data
from the California housing data set.
Returns:
A DataFrame that contains the features to be used for the model, including
synthetic features.
"""
selected_features = california_housing_dataframe[
["latitude",
"longitude",
"housing_median_age",
"total_rooms",
"total_bedrooms",
"population",
"households",
"median_income"]]
processed_features = selected_features.copy()
# Create a synthetic feature.
processed_features["rooms_per_person"] = (
california_housing_dataframe["total_rooms"] /
california_housing_dataframe["population"])
return processed_features
def preprocess_targets(california_housing_dataframe):
"""Prepares target features (i.e., labels) from California housing data set.
Args:
california_housing_dataframe: A Pandas DataFrame expected to contain data
from the California housing data set.
Returns:
A DataFrame that contains the target feature.
"""
output_targets = pd.DataFrame()
# Create a boolean categorical feature representing whether the
# median_house_value is above a set threshold.
output_targets["median_house_value_is_high"] = (
california_housing_dataframe["median_house_value"] > 265000).astype(float)
return output_targets
In [0]:
# Choose the first 12000 (out of 17000) examples for training.
training_examples = preprocess_features(california_housing_dataframe.head(12000))
training_targets = preprocess_targets(california_housing_dataframe.head(12000))
# Choose the last 5000 (out of 17000) examples for validation.
validation_examples = preprocess_features(california_housing_dataframe.tail(5000))
validation_targets = preprocess_targets(california_housing_dataframe.tail(5000))
# Double-check that we've done the right thing.
print("Training examples summary:")
display.display(training_examples.describe())
print("Validation examples summary:")
display.display(validation_examples.describe())
print("Training targets summary:")
display.display(training_targets.describe())
print("Validation targets summary:")
display.display(validation_targets.describe())
Para ver por qué es eficaz la regresión logística, primero entrenemos un modelo simple que use regresión lineal. Este modelo usará etiquetas con valores del conjunto {0, 1} e intentará predecir un valor continuo que esté lo más cerca posible de 0 o 1. Además, queremos interpretar el resultado como una probabilidad, de manera que sería ideal si el resultado estuviera dentro del rango (0, 1). De este modo, aplicaríamos un umbral de 0.5 para determinar la etiqueta.
Ejecuta las celdas a continuación para entrenar el modelo de regresión lineal con LinearRegressor.
In [0]:
def construct_feature_columns(input_features):
"""Construct the TensorFlow Feature Columns.
Args:
input_features: The names of the numerical input features to use.
Returns:
A set of feature columns
"""
return set([tf.feature_column.numeric_column(my_feature)
for my_feature in input_features])
In [0]:
def my_input_fn(features, targets, batch_size=1, shuffle=True, num_epochs=None):
"""Trains a linear regression model.
Args:
features: pandas DataFrame of features
targets: pandas DataFrame of targets
batch_size: Size of batches to be passed to the model
shuffle: True or False. Whether to shuffle the data.
num_epochs: Number of epochs for which data should be repeated. None = repeat indefinitely
Returns:
Tuple of (features, labels) for next data batch
"""
# Convert pandas data into a dict of np arrays.
features = {key:np.array(value) for key,value in dict(features).items()}
# Construct a dataset, and configure batching/repeating.
ds = Dataset.from_tensor_slices((features,targets)) # warning: 2GB limit
ds = ds.batch(batch_size).repeat(num_epochs)
# Shuffle the data, if specified.
if shuffle:
ds = ds.shuffle(10000)
# Return the next batch of data.
features, labels = ds.make_one_shot_iterator().get_next()
return features, labels
In [0]:
def train_linear_regressor_model(
learning_rate,
steps,
batch_size,
training_examples,
training_targets,
validation_examples,
validation_targets):
"""Trains a linear regression model.
In addition to training, this function also prints training progress information,
as well as a plot of the training and validation loss over time.
Args:
learning_rate: A `float`, the learning rate.
steps: A non-zero `int`, the total number of training steps. A training step
consists of a forward and backward pass using a single batch.
batch_size: A non-zero `int`, the batch size.
training_examples: A `DataFrame` containing one or more columns from
`california_housing_dataframe` to use as input features for training.
training_targets: A `DataFrame` containing exactly one column from
`california_housing_dataframe` to use as target for training.
validation_examples: A `DataFrame` containing one or more columns from
`california_housing_dataframe` to use as input features for validation.
validation_targets: A `DataFrame` containing exactly one column from
`california_housing_dataframe` to use as target for validation.
Returns:
A `LinearRegressor` object trained on the training data.
"""
periods = 10
steps_per_period = steps / periods
# Create a linear regressor object.
my_optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
my_optimizer = tf.contrib.estimator.clip_gradients_by_norm(my_optimizer, 5.0)
linear_regressor = tf.estimator.LinearRegressor(
feature_columns=construct_feature_columns(training_examples),
optimizer=my_optimizer
)
# Create input functions.
training_input_fn = lambda: my_input_fn(training_examples,
training_targets["median_house_value_is_high"],
batch_size=batch_size)
predict_training_input_fn = lambda: my_input_fn(training_examples,
training_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
predict_validation_input_fn = lambda: my_input_fn(validation_examples,
validation_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
# Train the model, but do so inside a loop so that we can periodically assess
# loss metrics.
print("Training model...")
print("RMSE (on training data):")
training_rmse = []
validation_rmse = []
for period in range (0, periods):
# Train the model, starting from the prior state.
linear_regressor.train(
input_fn=training_input_fn,
steps=steps_per_period
)
# Take a break and compute predictions.
training_predictions = linear_regressor.predict(input_fn=predict_training_input_fn)
training_predictions = np.array([item['predictions'][0] for item in training_predictions])
validation_predictions = linear_regressor.predict(input_fn=predict_validation_input_fn)
validation_predictions = np.array([item['predictions'][0] for item in validation_predictions])
# Compute training and validation loss.
training_root_mean_squared_error = math.sqrt(
metrics.mean_squared_error(training_predictions, training_targets))
validation_root_mean_squared_error = math.sqrt(
metrics.mean_squared_error(validation_predictions, validation_targets))
# Occasionally print the current loss.
print(" period %02d : %0.2f" % (period, training_root_mean_squared_error))
# Add the loss metrics from this period to our list.
training_rmse.append(training_root_mean_squared_error)
validation_rmse.append(validation_root_mean_squared_error)
print("Model training finished.")
# Output a graph of loss metrics over periods.
plt.ylabel("RMSE")
plt.xlabel("Periods")
plt.title("Root Mean Squared Error vs. Periods")
plt.tight_layout()
plt.plot(training_rmse, label="training")
plt.plot(validation_rmse, label="validation")
plt.legend()
return linear_regressor
In [0]:
linear_regressor = train_linear_regressor_model(
learning_rate=0.000001,
steps=200,
batch_size=20,
training_examples=training_examples,
training_targets=training_targets,
validation_examples=validation_examples,
validation_targets=validation_targets)
Examina las predicciones y decide si podemos usarlas o no para calcular la pérdida logística.
LinearRegressor usa la pérdida L2, que no se desempeña muy bien al penalizar las clasificaciones incorrectas cuando el resultado se interpreta como una probabilidad. Por ejemplo, debería haber una diferencia enorme si un ejemplo negativo se clasificara como positivo con una probabilidad de 0.9 frente a 0.9999, pero la pérdida L2 no diferencia estos casos de forma taxativa.
Por el contrario, LogLoss penaliza mucho más estos "errores de certeza". Recuerda que LogLoss se define de la siguiente manera:
Pero, primero, tendremos que obtener los valores de predicción. Podríamos usar LinearRegressor.predict para obtenerlos.
Dadas las predicciones y los objetivos, ¿podemos calcular LogLoss?
In [0]:
predict_validation_input_fn = lambda: my_input_fn(validation_examples,
validation_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
validation_predictions = linear_regressor.predict(input_fn=predict_validation_input_fn)
validation_predictions = np.array([item['predictions'][0] for item in validation_predictions])
_ = plt.hist(validation_predictions)
Para usar regresión logística, simplemente usa LinearClassifier en lugar de LinearRegressor. Completa el código a continuación.
NOTA: Al ejecutar train() y predict() en un modelo de LinearClassifier, puedes acceder a las probabilidades predichas con valores reales a través de la clave "probabilities" en el diccionario devuelto, p. ej., predictions["probabilities"]. La función log_loss de Sklearn es práctica para calcular la pérdida logística con estas probabilidades.
In [0]:
def train_linear_classifier_model(
learning_rate,
steps,
batch_size,
training_examples,
training_targets,
validation_examples,
validation_targets):
"""Trains a linear classification model.
In addition to training, this function also prints training progress information,
as well as a plot of the training and validation loss over time.
Args:
learning_rate: A `float`, the learning rate.
steps: A non-zero `int`, the total number of training steps. A training step
consists of a forward and backward pass using a single batch.
batch_size: A non-zero `int`, the batch size.
training_examples: A `DataFrame` containing one or more columns from
`california_housing_dataframe` to use as input features for training.
training_targets: A `DataFrame` containing exactly one column from
`california_housing_dataframe` to use as target for training.
validation_examples: A `DataFrame` containing one or more columns from
`california_housing_dataframe` to use as input features for validation.
validation_targets: A `DataFrame` containing exactly one column from
`california_housing_dataframe` to use as target for validation.
Returns:
A `LinearClassifier` object trained on the training data.
"""
periods = 10
steps_per_period = steps / periods
# Create a linear classifier object.
my_optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
my_optimizer = tf.contrib.estimator.clip_gradients_by_norm(my_optimizer, 5.0)
linear_classifier = # YOUR CODE HERE: Construct the linear classifier.
# Create input functions.
training_input_fn = lambda: my_input_fn(training_examples,
training_targets["median_house_value_is_high"],
batch_size=batch_size)
predict_training_input_fn = lambda: my_input_fn(training_examples,
training_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
predict_validation_input_fn = lambda: my_input_fn(validation_examples,
validation_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
# Train the model, but do so inside a loop so that we can periodically assess
# loss metrics.
print("Training model...")
print("LogLoss (on training data):")
training_log_losses = []
validation_log_losses = []
for period in range (0, periods):
# Train the model, starting from the prior state.
linear_classifier.train(
input_fn=training_input_fn,
steps=steps_per_period
)
# Take a break and compute predictions.
training_probabilities = linear_classifier.predict(input_fn=predict_training_input_fn)
training_probabilities = np.array([item['probabilities'] for item in training_probabilities])
validation_probabilities = linear_classifier.predict(input_fn=predict_validation_input_fn)
validation_probabilities = np.array([item['probabilities'] for item in validation_probabilities])
training_log_loss = metrics.log_loss(training_targets, training_probabilities)
validation_log_loss = metrics.log_loss(validation_targets, validation_probabilities)
# Occasionally print the current loss.
print(" period %02d : %0.2f" % (period, training_log_loss))
# Add the loss metrics from this period to our list.
training_log_losses.append(training_log_loss)
validation_log_losses.append(validation_log_loss)
print("Model training finished.")
# Output a graph of loss metrics over periods.
plt.ylabel("LogLoss")
plt.xlabel("Periods")
plt.title("LogLoss vs. Periods")
plt.tight_layout()
plt.plot(training_log_losses, label="training")
plt.plot(validation_log_losses, label="validation")
plt.legend()
return linear_classifier
In [0]:
linear_classifier = train_linear_classifier_model(
learning_rate=0.000005,
steps=500,
batch_size=20,
training_examples=training_examples,
training_targets=training_targets,
validation_examples=validation_examples,
validation_targets=validation_targets)
In [0]:
def train_linear_classifier_model(
learning_rate,
steps,
batch_size,
training_examples,
training_targets,
validation_examples,
validation_targets):
"""Trains a linear classification model.
In addition to training, this function also prints training progress information,
as well as a plot of the training and validation loss over time.
Args:
learning_rate: A `float`, the learning rate.
steps: A non-zero `int`, the total number of training steps. A training step
consists of a forward and backward pass using a single batch.
batch_size: A non-zero `int`, the batch size.
training_examples: A `DataFrame` containing one or more columns from
`california_housing_dataframe` to use as input features for training.
training_targets: A `DataFrame` containing exactly one column from
`california_housing_dataframe` to use as target for training.
validation_examples: A `DataFrame` containing one or more columns from
`california_housing_dataframe` to use as input features for validation.
validation_targets: A `DataFrame` containing exactly one column from
`california_housing_dataframe` to use as target for validation.
Returns:
A `LinearClassifier` object trained on the training data.
"""
periods = 10
steps_per_period = steps / periods
# Create a linear classifier object.
my_optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
my_optimizer = tf.contrib.estimator.clip_gradients_by_norm(my_optimizer, 5.0)
linear_classifier = tf.estimator.LinearClassifier(
feature_columns=construct_feature_columns(training_examples),
optimizer=my_optimizer
)
# Create input functions.
training_input_fn = lambda: my_input_fn(training_examples,
training_targets["median_house_value_is_high"],
batch_size=batch_size)
predict_training_input_fn = lambda: my_input_fn(training_examples,
training_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
predict_validation_input_fn = lambda: my_input_fn(validation_examples,
validation_targets["median_house_value_is_high"],
num_epochs=1,
shuffle=False)
# Train the model, but do so inside a loop so that we can periodically assess
# loss metrics.
print("Training model...")
print("LogLoss (on training data):")
training_log_losses = []
validation_log_losses = []
for period in range (0, periods):
# Train the model, starting from the prior state.
linear_classifier.train(
input_fn=training_input_fn,
steps=steps_per_period
)
# Take a break and compute predictions.
training_probabilities = linear_classifier.predict(input_fn=predict_training_input_fn)
training_probabilities = np.array([item['probabilities'] for item in training_probabilities])
validation_probabilities = linear_classifier.predict(input_fn=predict_validation_input_fn)
validation_probabilities = np.array([item['probabilities'] for item in validation_probabilities])
training_log_loss = metrics.log_loss(training_targets, training_probabilities)
validation_log_loss = metrics.log_loss(validation_targets, validation_probabilities)
# Occasionally print the current loss.
print(" period %02d : %0.2f" % (period, training_log_loss))
# Add the loss metrics from this period to our list.
training_log_losses.append(training_log_loss)
validation_log_losses.append(validation_log_loss)
print("Model training finished.")
# Output a graph of loss metrics over periods.
plt.ylabel("LogLoss")
plt.xlabel("Periods")
plt.title("LogLoss vs. Periods")
plt.tight_layout()
plt.plot(training_log_losses, label="training")
plt.plot(validation_log_losses, label="validation")
plt.legend()
return linear_classifier
In [0]:
linear_classifier = train_linear_classifier_model(
learning_rate=0.000005,
steps=500,
batch_size=20,
training_examples=training_examples,
training_targets=training_targets,
validation_examples=validation_examples,
validation_targets=validation_targets)
Algunas de las métricas que resultan útiles para la clasificación son la exactitud del modelo, la curva ROC y el área bajo la curva ROC (AUC). Examinaremos estas métricas.
LinearClassifier.evaluate calcula las métricas útiles, como la exactitud y el AUC.
In [0]:
evaluation_metrics = linear_classifier.evaluate(input_fn=predict_validation_input_fn)
print("AUC on the validation set: %0.2f" % evaluation_metrics['auc'])
print("Accuracy on the validation set: %0.2f" % evaluation_metrics['accuracy'])
Para obtener las tasas de verdaderos positivos y falsos positivos que se necesitan para representar una curva ROC, puedes usar probabilidades de clase, como aquellas que se calculan con LinearClassifier.predict, y la función roc_curve de Sklearn.
In [0]:
validation_probabilities = linear_classifier.predict(input_fn=predict_validation_input_fn)
# Get just the probabilities for the positive class.
validation_probabilities = np.array([item['probabilities'][1] for item in validation_probabilities])
false_positive_rate, true_positive_rate, thresholds = metrics.roc_curve(
validation_targets, validation_probabilities)
plt.plot(false_positive_rate, true_positive_rate, label="our model")
plt.plot([0, 1], [0, 1], label="random classifier")
_ = plt.legend(loc=2)
Comprueba si puedes ajustar la configuración de aprendizaje del modelo entrenado en la Tarea 2 para mejorar el AUC.
Con frecuencia, determinadas métricas mejoran en detrimento de otras, y debes encontrar la configuración que logre un buen equilibrio.
Verifica si todas las métricas mejoran al mismo tiempo.
In [0]:
# TUNE THE SETTINGS BELOW TO IMPROVE AUC
linear_classifier = train_linear_classifier_model(
learning_rate=0.000005,
steps=500,
batch_size=20,
training_examples=training_examples,
training_targets=training_targets,
validation_examples=validation_examples,
validation_targets=validation_targets)
evaluation_metrics = linear_classifier.evaluate(input_fn=predict_validation_input_fn)
print("AUC on the validation set: %0.2f" % evaluation_metrics['auc'])
print("Accuracy on the validation set: %0.2f" % evaluation_metrics['accuracy'])
Una solución posible que funciona es simplemente entrenar por más tiempo, siempre que no realicemos un sobreajuste.
Esto se puede lograr al incrementar el número de pasos, el tamaño del lote o ambos.
Todas las métricas mejoran a la vez, de manera que nuestra métrica de pérdida es un buen representante tanto del AUC como de la exactitud.
Observa cómo hace muchas más iteraciones simplemente para restringir algunas unidades más del AUC. Esto ocurre comúnmente, pero, con frecuencia, incluso esta pequeña ganancia vale la pena.
In [0]:
linear_classifier = train_linear_classifier_model(
learning_rate=0.000003,
steps=20000,
batch_size=500,
training_examples=training_examples,
training_targets=training_targets,
validation_examples=validation_examples,
validation_targets=validation_targets)
evaluation_metrics = linear_classifier.evaluate(input_fn=predict_validation_input_fn)
print("AUC on the validation set: %0.2f" % evaluation_metrics['auc'])
print("Accuracy on the validation set: %0.2f" % evaluation_metrics['accuracy'])