shell
pip install pylearn2
python
import numpy as np
from pylearn2.datasets import DenseDesignMatrix
X_train = np.load('train_data.npy')
y_train = np.load('train_labels.npy')
X_test = np.load('test_data.npy')
y_test = np.load('test_labels.npy')
train_set = DenseDesignMatrix(X=X_train, y=y_train)
test_set = DenseDesignMatrix(X=X_test, y=y_test)
python
from pylearn2.models import mlp
from pylearn2.training_algorithms import sgd
from pylearn2.termination_criteria import EpochCounter
input_dim = X_train.shape[1]
output_dim = np.max(y_train) + 1
hidden_dim = 100
model = mlp.MLP(
layers=[
mlp.Sigmoid(layer_name='hidden', dim=hidden_dim),
mlp.Softmax(layer_name='output', dim=output_dim)
],
nvis=input_dim
)
learning_rate = 0.1
batch_size = 128
n_epochs = 100
algorithm = sgd.SGD(
learning_rate=learning_rate,
batch_size=batch_size,
termination_criterion=EpochCounter(max_epochs=n_epochs),
monitoring_dataset={'train': train_set, 'test': test_set}
)
model.train(dataset=train_set, algorithm=algorithm)
python
model.train(dataset=train_set, algorithm=algorithm)
test_error = model.eval(dataset=test_set)
python
new_data = np.load('new_data.npy')
predictions = model.predict(new_data)
