theano models

load and prepare the dataset

from loader import Loader
from trainer import Trainer

<!-- load dataset -->
load = Loader('/path/to/dataset/mnist')
trX, trY, teX, teY = load.mnist()
dataset = (trX, trY, teX, teY)
trainer = Trainer(dataset)

training logistic regression

trainer.train_logistic_regression(distribution='randn', fan_in=28*28, fan_out=10)

train multilayer perceptron

trainer.train_multilayer_perceptron(distribution='randn', fan_in=28*28, n_hidden=500, fan_out=10)

train autoencoder

trainer.train_autoencoder(distribution='randn', fan_in=28*28, fan_out=28*28)

train denoisy autoencoder

trainer.train_denoisy_autoencoder(distribution='randn', n_visible=28*28, n_hidden=500)

train stacked denoisy autoencoder

trainer.train_stacked_denoisy_autoencoder(distribution='randn',
fan_in=784, n_hidden_sizes=[500, 1024], fan_out=10, noise_levels=[0.1,
0.2])

train contractive autoencoder

trainer.train_contractive_autoencoder(distribution='randn', fan_in=784, fan_out=10)

train restricted boltzmann machine

trainer.train_rbm(distribution='randn')