Last active
September 15, 2017 23:52
-
-
Save bacalfa/c8a37feb7e9443f0fc5797bd29c797b5 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import numpy as np | |
from sklearn.neural_network import MLPRegressor | |
from sklearn.model_selection import RandomizedSearchCV | |
from keras.models import Sequential | |
from keras.layers import Dense | |
from keras.wrappers.scikit_learn import KerasRegressor | |
import sklearn.metrics | |
def mean_absolute_percentage_error(y_true, y_pred): | |
if any(y_true == 0): | |
return 0.0 | |
else: | |
return np.mean(np.abs((y_true - y_pred) / y_true)) * 100 | |
import time | |
## Run on CPU only | |
import os | |
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" # see issue #152 | |
os.environ["CUDA_VISIBLE_DEVICES"] = "" | |
## Fix random seed for reproducibility | |
seed = 7 | |
np.random.seed(seed) | |
## General and CV options | |
n_iter = 80 | |
n_jobs = -1 | |
cv = 5 | |
max_iter = 1000 | |
batch_size = 32 | |
## Import data | |
classification = False | |
X = np.loadtxt("house_dataset_x.txt") | |
y = np.loadtxt("house_dataset_y.txt") | |
from sklearn.model_selection import train_test_split | |
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=seed) | |
from sklearn import preprocessing | |
X_train, X_test = preprocessing.scale(X_train), preprocessing.scale(X_test) | |
# X_train, X_test = preprocessing.minmax_scale(X_train), preprocessing.minmax_scale(X_test) | |
## Parameters for MLP | |
neurons_output = max(1, y_train.shape[1] if len(y_train.shape) > 1 else 1) | |
sumInOutSize = X_train.shape[1] + neurons_output | |
minNeuron = neurons_output | |
maxNeuron = int(X_train.shape[0] / (2.0 * (X_train.shape[1] + neurons_output))) | |
epochs = np.ceil(np.float(batch_size)/X_train.shape[0]*max_iter).astype(np.int_) | |
input_dim = X_train.shape[1] if len(X_train.shape) > 1 else 1 | |
output_dim = y_train.shape[1] if len(y_train.shape) > 1 else 1 | |
n_layers = 1 | |
## MLPRegressor | |
MLPR_mdl = MLPRegressor(random_state=seed, max_iter=max_iter, batch_size=batch_size) | |
MLPR_param_distributions = \ | |
{ | |
"hidden_layer_sizes": [(x,) for x in range(minNeuron,maxNeuron)], | |
"solver": ["lbfgs", "sgd", "adam"], | |
"learning_rate": ["constant", "invscaling", "adaptive"], | |
"activation": ["identity", "logistic", "tanh", "relu"], | |
"alpha": list(np.random.uniform(low=0, high=5, size=n_iter)), | |
"learning_rate_init": list(10.0 ** np.random.uniform(low=-6, high=-2, size=n_iter)), | |
"momentum": list(np.random.uniform(low=1E-2, high=1, size=n_iter)), | |
"nesterovs_momentum": [True, False], | |
"beta_1": list(np.random.uniform(low=1E-2, high=1, size=n_iter)), | |
"beta_2": list(np.random.uniform(low=1E-2, high=1, size=n_iter)) | |
} | |
MLPR_random_search = RandomizedSearchCV(estimator=MLPR_mdl, cv=cv, n_iter=n_iter, param_distributions=MLPR_param_distributions, n_jobs=n_jobs) | |
MLPR_start = time.perf_counter() | |
MLPR_result = MLPR_random_search.fit(X_train, y_train) | |
MLPR_end = time.perf_counter() | |
MLPR_y_pred = MLPR_result.predict(X=X_test) | |
## KerasRegressor | |
# Function to create model, required for KerasClassifier | |
def create_model(n_layers=n_layers, units=10, input_dim=input_dim, output_dim=output_dim, | |
optimizer="rmsprop", loss="binary_crossentropy", | |
kernel_initializer="glorot_uniform", activation="sigmoid", | |
kernel_regularizer="l2", kernel_regularizer_weight=0.01, | |
lr=0.01, momentum=0.0, decay=0.0, nesterov=False, rho=0.9, epsilon=1E-8, | |
beta_1=0.9, beta_2=0.999, schedule_decay=0.004): | |
from keras import regularizers, optimizers | |
# Create model | |
if kernel_regularizer.lower() == "l1": | |
kernel_regularizer = regularizers.l1(l=kernel_regularizer_weight) | |
elif kernel_regularizer.lower() == "l2": | |
kernel_regularizer = regularizers.l2(l=kernel_regularizer_weight) | |
elif kernel_regularizer.lower() == "l1_l2": | |
kernel_regularizer = regularizers.l1_l2(l1=kernel_regularizer_weight, l2=kernel_regularizer_weight) | |
else: | |
print("Warning: Kernel regularizer {0} not supported. Using default 'l2' regularizer.".format( | |
kernel_regularizer)) | |
kernel_regularizer = regularizers.l2(l=kernel_regularizer_weight) | |
if optimizer.lower() == "sgd": | |
optimizer = optimizers.sgd(lr=lr, momentum=momentum, decay=decay, nesterov=nesterov) | |
elif optimizer.lower() == "rmsprop": | |
optimizer = optimizers.rmsprop(lr=lr, rho=rho, epsilon=epsilon, decay=decay) | |
elif optimizer.lower() == "adagrad": | |
optimizer = optimizers.adagrad(lr=lr, epsilon=epsilon, decay=decay) | |
elif optimizer.lower() == "adadelta": | |
optimizer = optimizers.adadelta(lr=lr, rho=rho, epsilon=epsilon, decay=decay) | |
elif optimizer.lower() == "adam": | |
optimizer = optimizers.adam(lr=lr, beta_1=beta_1, beta_2=beta_2, epsilon=epsilon, decay=decay) | |
elif optimizer.lower() == "adamax": | |
optimizer = optimizers.adamax(lr=lr, beta_1=beta_1, beta_2=beta_2, epsilon=epsilon, decay=decay) | |
elif optimizer.lower() == "nadam": | |
optimizer = optimizers.nadam(lr=lr, beta_1=beta_1, beta_2=beta_2, epsilon=epsilon, | |
schedule_decay=schedule_decay) | |
else: | |
print("Warning: Optimizer {0} not supported. Using default 'sgd' optimizer.".format(optimizer)) | |
optimizer = "sgd" | |
model = Sequential() | |
model.add( | |
Dense(units=units, input_dim=input_dim, | |
kernel_initializer=kernel_initializer, activation=activation, | |
kernel_regularizer=kernel_regularizer)) | |
for layer_count in range(n_layers - 1): | |
model.add( | |
Dense(units=units, kernel_initializer=kernel_initializer, activation=activation, | |
kernel_regularizer=kernel_regularizer)) | |
model.add(Dense(units=output_dim, | |
kernel_initializer=kernel_initializer, activation="linear", | |
kernel_regularizer=kernel_regularizer)) | |
# Compile model | |
model.compile(loss=loss, optimizer=optimizer, metrics=['accuracy']) | |
return model | |
keras_param_distributions = \ | |
{ | |
"units": [x for x in range(minNeuron,maxNeuron)], | |
"optimizer": ["sgd", "rmsprop", "adagrad", "adadelta", "adam", "adamax", "nadam"], | |
"kernel_initializer": ["uniform", "normal", "glorot_normal", "glorot_uniform"], | |
"activation": ["relu", "tanh", "sigmoid"], | |
"loss": ["mse"], | |
"kernel_regularizer": ["l1", "l2", "l1_l2"], | |
"kernel_regularizer_weight": list(np.random.uniform(low=1E-3, high=5, size=n_iter)), | |
"lr": list(10.0 ** np.random.uniform(low=-6, high=0, size=n_iter)), | |
"momentum": list(np.random.uniform(low=1E-4, high=1, size=n_iter)), | |
# "decay": list(np.random.uniform(low=1E-8, high=1E-2, size=n_iter)), | |
"nesterov": [True, False], | |
# "rho": list(np.random.uniform(low=1E-1, high=1, size=n_iter)), | |
# "epsilon": list(np.random.uniform(low=1E-8, high=1E-6, size=n_iter)), | |
# "schedule_decay": list(np.random.uniform(low=1E-4, high=1E-2, size=n_iter)), | |
"beta_1": list(np.random.uniform(low=1E-2, high=1, size=n_iter)), | |
"beta_2": list(np.random.uniform(low=1E-2, high=1, size=n_iter)) | |
} | |
keras_mdl = KerasRegressor(build_fn=create_model, epochs=epochs, batch_size=batch_size, verbose=0) | |
keras_random_search = RandomizedSearchCV(estimator=keras_mdl, cv=cv, n_iter=n_iter, param_distributions=keras_param_distributions, n_jobs=n_jobs) | |
keras_start = time.perf_counter() | |
keras_result = keras_random_search.fit(X_train, y_train) | |
keras_end = time.perf_counter() | |
keras_y_pred = keras_result.predict(X=X_test) | |
## Summarize results | |
print("MLPRegressor") | |
print("============") | |
print("Time: {0}".format(MLPR_end - MLPR_start)) | |
print("Score: {0}".format(MLPR_result.best_score_)) | |
print("Parameters: {0}".format(MLPR_result.best_params_)) | |
print("r2_score: {0}".format(sklearn.metrics.r2_score(y_true=y_test, y_pred=MLPR_y_pred))) | |
print("mean_absolute_error: {0}".format(sklearn.metrics.mean_absolute_error(y_true=y_test, y_pred=MLPR_y_pred))) | |
print("mean_squared_error: {0}".format(sklearn.metrics.mean_squared_error(y_true=y_test, y_pred=MLPR_y_pred))) | |
print("median_absolute_error: {0}".format(sklearn.metrics.median_absolute_error(y_true=y_test, y_pred=MLPR_y_pred))) | |
print("mean_absolute_percentage_error: {0}".format(mean_absolute_percentage_error(y_true=y_test, y_pred=MLPR_y_pred))) | |
print() | |
print("KerasRegressor") | |
print("==============") | |
print("Time: {0}".format(keras_end - keras_start)) | |
print("Score: {0}".format(keras_result.best_score_)) | |
print("Parameters: {0}".format(keras_result.best_params_)) | |
print("r2_score: {0}".format(sklearn.metrics.r2_score(y_true=y_test, y_pred=keras_y_pred))) | |
print("mean_absolute_error: {0}".format(sklearn.metrics.mean_absolute_error(y_true=y_test, y_pred=keras_y_pred))) | |
print("mean_squared_error: {0}".format(sklearn.metrics.mean_squared_error(y_true=y_test, y_pred=keras_y_pred))) | |
print("median_absolute_error: {0}".format(sklearn.metrics.median_absolute_error(y_true=y_test, y_pred=keras_y_pred))) | |
print("mean_absolute_percentage_error: {0}".format(mean_absolute_percentage_error(y_true=y_test, y_pred=keras_y_pred))) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment