Package pylearn :: Package sandbox :: Module denoising_aa
[hide private]

Module denoising_aa

source code

A denoising auto-encoder


Warning: You should use this interface. It is not complete and is not functional. Instead, use:

   ssh://projects@lgcm.iro.umontreal.ca/repos/denoising_aa

Classes [hide private]
DenoisingAutoEncoder
DenoisingAutoEncoderModel
Functions [hide private]
 
hiding_corruption_formula(seed, average_fraction_hidden)
Return a formula for the corruption process, in which a random subset of the input numbers are hidden (mapped to 0).
source code
 
squash_affine_formula(squash_function=sigmoid)
Simply does: squash_function(b + xW) By convention prefix the parameters by _
source code
 
gradient_descent_update_formula() source code
 
probabilistic_classifier_loss_formula() source code
 
binomial_cross_entropy_formula() source code
 
squash_affine_autoencoder_formula(hidden_squash=t.tanh, reconstruction_squash=sigmoid, share_weights=True, reconstruction_nll_formula=binomial_cross_entropy_formula(), update_formula=gradient_descent_update_formula) source code

Imports: theano, t, math, binomial


Function Details [hide private]

hiding_corruption_formula(seed, average_fraction_hidden)

source code 

Return a formula for the corruption process, in which a random subset of the input numbers are hidden (mapped to 0).

Parameters:
  • seed (anything that numpy.random.RandomState accepts) - seed of the random generator
  • average_fraction_hidden (0 <= real number <= 1) - the probability with which each input number is hidden (set to 0).