Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit history

Commits on Jan 31, 2013

• Use list of pairs for updates, not dict, to remove warning

  lamblin

  committed
  0440226

  Copy full SHA for 0440226

Commits on Oct 31, 2012

• make the number of hidden unit for mlp parametrisable.

  nouiz

  committed
  eb315c6

  Copy full SHA for eb315c6

• small formating change

  nouiz

  committed
  7ef0549

  Copy full SHA for 7ef0549

Commits on Sep 6, 2012

• use borrow=True when making shared variable to make the example work with 32 bits python.

  nouiz

  committed
  8596e69

  Copy full SHA for 8596e69

Commits on Jun 16, 2012

• update the best_iter variable in the mlp.py example.

  nouiz

  committed
  bb21f3e

  Copy full SHA for bb21f3e

Commits on Mar 20, 2012

• indent.

  nouiz

  committed
  362a8c0

  Copy full SHA for 362a8c0

• pep8

  nouiz

  committed
  da46247

  Copy full SHA for da46247

• remove trailing white space.

  nouiz

  committed
  69d7c41

  Copy full SHA for 69d7c41

Commits on Dec 8, 2011

• mlp.HiddenLayer now accepts parameters W and b by reference

  Show description for 5bc54bd
  gdesjardins

  committed
  5bc54bd

  Copy full SHA for 5bc54bd

Commits on Jul 17, 2011

• Use numpy.inf instead of float('inf').

  Show description for 230e17f
  lamblin

  committed
  230e17f

  Copy full SHA for 230e17f

Commits on Mar 15, 2011

• Remove use of .value in mlp

  lamblin

  committed
  46fc8c9

  Copy full SHA for 46fc8c9

Commits on Jul 22, 2010

• fix syntax

  nouiz

  committed
  d97f381

  Copy full SHA for d97f381

Commits on Jul 19, 2010

• added comment.

  nouiz

  committed
  d3ab267

  Copy full SHA for d3ab267

• Merge git://github.com/fidlej/DeepLearningTutorials

  nouiz

  committed
  d7ea6c2

  Copy full SHA for d7ea6c2

Commits on Jul 16, 2010

• fix type and doc formating.

  nouiz

  committed
  e371d52

  Copy full SHA for e371d52

• Unified the W_values initialization.

  fidlej

  committed
  8469cc9

  Copy full SHA for 8469cc9

Commits on Apr 1, 2010

• removed the expected time from the tutorial code. Now into test.py:speed

  nouiz

  committed
  a8ee2e6

  Copy full SHA for a8ee2e6

Commits on Mar 26, 2010

• merging

  

  Razvan Pascanu

  committed
  a59a2bf

  Copy full SHA for a59a2bf

Commits on Mar 19, 2010

• moved the timing into the code to time only the exec time.

  nouiz

  committed
  223e870

  Copy full SHA for 223e870

Commits on Mar 16, 2010

• first draft DBN + sqrt(6/(n_in+n_out)) + different initializations for tanh/sigmoid

  

  Razvan Pascanu

  committed
  1172608

  Copy full SHA for 1172608

Commits on Feb 19, 2010

• I merged my version 2 into the official deep learning tutorials

  

  Razvan Pascanu

  committed
  d71dce9

  Copy full SHA for d71dce9

Commits on Feb 12, 2010

• a few typos + a few updates to rbm

  

  Razvan Pascanu

  committed
  cffce58

  Copy full SHA for cffce58

Commits on Feb 10, 2010

• added links to the stacked denoising autoencoder tutorial

  

  Razvan Pascanu

  committed
  dffd751

  Copy full SHA for dffd751

Commits on Feb 8, 2010

• working version of SdA.py

  

  Razvan Pascanu

  committed
  c8fc56d

  Copy full SHA for c8fc56d

Commits on Feb 5, 2010

• Using index instead of offset ; some bug fixes for the SdA.py; partially finished SdA tutorial ( I got stucked in trying to fix the code and didn't manage to finish everything )

  

  Razvan Pascanu

  committed
  9d2f5d3

  Copy full SHA for 9d2f5d3

Commits on Feb 3, 2010

• I have made all tutorials (code and documentation) to use shared variable to store dataset as James suggested. I also changed the validation_frequency to Yoshua's suggestion. One should still need…

  Show description for a074b32
  pascanur

  committed
  a074b32

  Copy full SHA for a074b32

Commits on Feb 2, 2010

• Changing to the new way of storing the dataset (using shared variables)

  pascanur

  committed
  ffa5bf9

  Copy full SHA for ffa5bf9

Commits on Jan 28, 2010

• Fixed signed bug in weight initialization

  pascanur

  committed
  4256877

  Copy full SHA for 4256877

Commits on Jan 22, 2010

• latest updates to convolutional code

  Show description for e0e5e5d
  gdesjardins

  committed
  e0e5e5d

  Copy full SHA for e0e5e5d

Commits on Jan 21, 2010

• SdA.py implements a stacked de-noising autoencoder in the tutorial style. I think DBN/SdA are a bit different from the rest, so if anybody has the time to read through it and give some feedback I w…

  Show description for 76f20c3
  pascanur

  committed
  76f20c3

  Copy full SHA for 76f20c3

Commits on Jan 19, 2010

• I redone the tests and updated final scores and how long they run

  pascanur

  committed
  61ceb18

  Copy full SHA for 61ceb18

Commits on Jan 17, 2010

• Change 'T.sum' to 'T.mean' in computations of NLL, change the documentation accordingly

  

  Pascal Lamblin

  committed
  3ed2a7f

  Copy full SHA for 3ed2a7f

Commits on Jan 15, 2010

• switching to T.sum when computing negative log likelihood (though T.mean should also be supported) for mlp and logistic_sgd. Second bug fixed was when initializing the weights, we need 6./(n_in+n_o…

  Show description for ff4bd33
  pascanur

  committed
  ff4bd33

  Copy full SHA for ff4bd33

• fixed typo breaking the execution

  

  Yoshua Bengio

  committed
  a8926d3

  Copy full SHA for a8926d3

• minor doc changes + initialization in sqrt(6/(fan_in+fan_out)) instead of sqrt(1/fan_in)

  

  Yoshua Bengio

  committed
  df26db2

  Copy full SHA for df26db2