Fri, 11 Jul 2008 15:33:27 -0400 |
Joseph Turian |
Small bugfix
|
Fri, 11 Jul 2008 15:29:37 -0400 |
Joseph Turian |
Added noise functions
|
Fri, 11 Jul 2008 15:26:41 -0400 |
Joseph Turian |
Added cost functions
|
Fri, 11 Jul 2008 15:14:03 -0400 |
Joseph Turian |
Updated readme
|
Fri, 11 Jul 2008 15:13:44 -0400 |
Joseph Turian |
Autoassociator now seems to work
|
Fri, 11 Jul 2008 01:33:27 -0400 |
Joseph Turian |
XENT per instance
|
Thu, 10 Jul 2008 17:33:28 -0400 |
Joseph Turian |
Moved denoising AA to sandbox
|
Mon, 14 Jul 2008 13:48:36 -0400 |
Frederic Bastien |
corrected import as a file moved to sandbox
|
Thu, 10 Jul 2008 09:03:11 -0400 |
Frederic Bastien |
Automated merge with ssh://projects@lgcm.iro.umontreal.ca/hg/pylearn
|
Thu, 10 Jul 2008 02:10:23 -0400 |
Joseph Turian |
Globals are now parameters for the RBM model
|
Thu, 10 Jul 2008 01:17:40 -0400 |
Joseph Turian |
Added weight decay
|
Thu, 10 Jul 2008 00:51:32 -0400 |
Joseph Turian |
Debugging simple AA a bit
|
Wed, 09 Jul 2008 17:55:46 -0400 |
Yoshua Bengio |
Making linear_regression more robust
|
Wed, 09 Jul 2008 15:27:12 -0400 |
Joseph Turian |
* Added momentum.
|
Wed, 09 Jul 2008 00:29:16 -0400 |
Joseph Turian |
Small updates
|
Tue, 08 Jul 2008 23:59:57 -0400 |
Joseph Turian |
Cleaned up, added sparse_instance
|
Tue, 08 Jul 2008 21:42:21 -0400 |
Joseph Turian |
RBM minibatch works
|
Tue, 08 Jul 2008 20:48:56 -0400 |
Joseph Turian |
Better output
|
Tue, 08 Jul 2008 20:39:36 -0400 |
Joseph Turian |
Added init to sandbox
|
Tue, 08 Jul 2008 20:37:36 -0400 |
Joseph Turian |
Bug fixed in RBM
|
Tue, 08 Jul 2008 20:14:21 -0400 |
Joseph Turian |
Basic, broken RBM implementation
|
Tue, 08 Jul 2008 18:46:37 -0400 |
Joseph Turian |
merge
|
Tue, 08 Jul 2008 18:46:26 -0400 |
Joseph Turian |
Made sandbox directory
|
Tue, 08 Jul 2008 18:45:35 -0400 |
Joseph Turian |
More debugging
|
Tue, 08 Jul 2008 17:50:23 -0400 |
Yoshua Bengio |
merge
|
Tue, 08 Jul 2008 17:49:44 -0400 |
Yoshua Bengio |
First non-crashing draft of LinearRegression
|
Tue, 08 Jul 2008 17:41:45 -0400 |
Joseph Turian |
Renamed simple AA directory
|
Tue, 08 Jul 2008 17:41:26 -0400 |
Joseph Turian |
Debugging simple AA
|
Tue, 08 Jul 2008 17:08:23 -0400 |
Joseph Turian |
__str__ method
|
Tue, 08 Jul 2008 02:27:00 -0400 |
Joseph Turian |
Added a simple AA
|
Tue, 08 Jul 2008 02:00:14 -0400 |
Joseph Turian |
merge
|
Tue, 08 Jul 2008 02:00:00 -0400 |
Joseph Turian |
Switched back to margin loss
|
Tue, 08 Jul 2008 01:59:42 -0400 |
Joseph Turian |
Small fix
|
Tue, 08 Jul 2008 01:58:16 -0400 |
Joseph Turian |
Moved xent loss to nnet_ups
|
Tue, 08 Jul 2008 01:57:33 -0400 |
Joseph Turian |
Added xent loss
|
Wed, 09 Jul 2008 16:55:27 -0400 |
Frederic Bastien |
added function to load amat file
|
Mon, 07 Jul 2008 12:27:06 -0400 |
Frederic Bastien |
small modif by yoshue
|
Mon, 07 Jul 2008 12:02:34 -0400 |
Yoshua Bengio |
fixing merge
|
Mon, 07 Jul 2008 10:09:37 -0400 |
Yoshua Bengio |
merge
|
Mon, 07 Jul 2008 10:08:35 -0400 |
Yoshua Bengio |
Redesigning linear_regression
|
Mon, 16 Jun 2008 17:47:36 -0400 |
Yoshua Bengio |
Automated merge with http://lgcm.iro.umontreal.ca/hg/pylearn
|
Tue, 03 Jun 2008 21:27:32 -0400 |
bengioy |
split LearningAlgorithm into OfflineLearningAlgorithm and OnlineLearningAlgorithm
|
Mon, 07 Jul 2008 02:15:14 -0400 |
Joseph Turian |
No update if there aren't any non-zeros
|
Mon, 07 Jul 2008 02:06:15 -0400 |
Joseph Turian |
Moved more logic into model.py
|
Mon, 07 Jul 2008 01:57:49 -0400 |
Joseph Turian |
Update doc
|
Mon, 07 Jul 2008 01:54:46 -0400 |
Joseph Turian |
Moved sparse_random_autoassociator from my repository
|
Fri, 04 Jul 2008 01:00:22 -0400 |
Joseph Turian |
Merge
|
Fri, 04 Jul 2008 01:00:11 -0400 |
Joseph Turian |
Removed featuremap.py, moved to common
|
Thu, 03 Jul 2008 23:55:39 -0400 |
Joseph Turian |
merge
|
Thu, 03 Jul 2008 23:52:56 -0400 |
Joseph Turian |
Removed common/
|
Thu, 03 Jul 2008 18:48:32 -0400 |
Joseph Turian |
Bugfix in common.misc.find_files
|
Thu, 03 Jul 2008 17:52:26 -0400 |
Joseph Turian |
merge
|
Thu, 03 Jul 2008 17:52:11 -0400 |
Joseph Turian |
Added to misc.file
|
Thu, 03 Jul 2008 00:14:42 -0400 |
Joseph Turian |
Added common.misc.utilsdir
|
Wed, 02 Jul 2008 23:56:51 -0400 |
Joseph Turian |
Added common.misc.homedir()
|
Wed, 02 Jul 2008 16:10:00 -0400 |
Joseph Turian |
common.misc.runcmd can now take input
|
Wed, 02 Jul 2008 15:42:42 -0400 |
Joseph Turian |
Improved runcmd
|
Tue, 01 Jul 2008 21:09:24 -0400 |
Joseph Turian |
Added common.misc.runcmd
|
Thu, 19 Jun 2008 16:27:02 -0400 |
Joseph Turian |
Added common.string
|
Thu, 19 Jun 2008 16:18:37 -0400 |
Joseph Turian |
Added more functions
|
Thu, 19 Jun 2008 16:12:29 -0400 |
Joseph Turian |
Added common directory
|
Thu, 19 Jun 2008 12:35:41 -0400 |
Thierry Bertin-Mahieux |
dataset__call__() returns a FieldsSubsetDataSet, so still a subset of fields, but not cached any more. I added the function dataset.cached_fields_subset(self,*fieldnames) that returns the old version, cached, in case someone needs it. Current behaviour passes the tests.
|
Tue, 17 Jun 2008 17:12:43 -0400 |
Thierry Bertin-Mahieux |
Cached dataset seems debug, using n_batches... is n_batches around to stay?
|
Tue, 17 Jun 2008 16:50:35 -0400 |
Frederic Bastien |
code cleanup
|
Tue, 17 Jun 2008 16:47:33 -0400 |
Frederic Bastien |
debugged RenamedFieldsDataSet
|
Tue, 17 Jun 2008 16:20:17 -0400 |
Frederic Bastien |
Can give test name in debug mode. This will execute on the gived test name.
|
Tue, 17 Jun 2008 16:09:59 -0400 |
Frederic Bastien |
added test for ds.subset()
|
Tue, 17 Jun 2008 14:59:10 -0400 |
Frederic Bastien |
corrected test_RenamedFieldsDataSet. It still fail, but this should be du to bug in RenamedFieldsDataSet.
|
Tue, 17 Jun 2008 14:46:15 -0400 |
Frederic Bastien |
Automated merge with ssh://projects@lgcm.iro.umontreal.ca/hg/pylearn
|
Tue, 17 Jun 2008 14:33:15 -0400 |
Yoshua Bengio |
Automated merge with ssh://projects@lgcm.iro.umontreal.ca/hg/pylearn
|
Tue, 17 Jun 2008 14:32:54 -0400 |
Yoshua Bengio |
Added test for dataset/RenamedFieldsDataSet
|
Tue, 17 Jun 2008 14:46:10 -0400 |
Frederic Bastien |
bugfix, MinibatchDataSet.minibatches_nowrap() must raise StopIteration when needed
|
Tue, 17 Jun 2008 14:45:22 -0400 |
Frederic Bastien |
added test for ds.subset
|
Tue, 17 Jun 2008 14:32:03 -0400 |
Frederic Bastien |
fix test for __getitem__
|
Tue, 17 Jun 2008 13:33:17 -0400 |
Frederic Bastien |
corrected test for dataset.minibatches()
|
Tue, 17 Jun 2008 11:41:01 -0400 |
Frederic Bastien |
Automated merge with ssh://projects@lgcm.iro.umontreal.ca/hg/pylearn
|
Mon, 16 Jun 2008 17:26:51 -0400 |
Thierry Bertin-Mahieux |
n_minibatches in ArrayDataSet automatically computed
|
Mon, 16 Jun 2008 17:17:50 -0400 |
Frederic Bastien |
Automated merge with ssh://projects@lgcm.iro.umontreal.ca/hg/pylearn
|
Mon, 16 Jun 2008 17:17:01 -0400 |
Thierry Bertin-Mahieux |
debugging when fieldnames = None, now -> fieldnames = fieldNames()
|
Mon, 16 Jun 2008 17:07:44 -0400 |
Frederic Bastien |
Automated merge with ssh://projects@lgcm.iro.umontreal.ca/hg/pylearn
|
Mon, 16 Jun 2008 16:59:35 -0400 |
Thierry Bertin-Mahieux |
sorry for all the debugging, this push should be the good one (and damn you *self*)
|
Mon, 16 Jun 2008 16:52:39 -0400 |
Thierry Bertin-Mahieux |
debug in an assert of minibatches
|
Mon, 16 Jun 2008 17:07:41 -0400 |
Frederic Bastien |
bugfix in RenamedDataSet
|
Mon, 16 Jun 2008 16:44:11 -0400 |
Thierry Bertin-Mahieux |
redone my previous hack in MinibatchToSingleExampleIterator, tests should work again
|
Mon, 16 Jun 2008 16:38:03 -0400 |
Thierry Bertin-Mahieux |
IMPORTANT: minibatches now returns minibatch_nowrap with a minimum of assert before. Should implement the good behavior, e.g. returning only complete batches and let the user figure out what he wants.
|
Mon, 16 Jun 2008 16:11:36 -0400 |
Thierry Bertin-Mahieux |
merge
|
Mon, 16 Jun 2008 16:03:16 -0400 |
Thierry Bertin-Mahieux |
little hack in MiniBatchToSingleExampleIterator, there was a problem which I think was not a bug, we were receiving [array(3)] and everything was crashing. Hack is kinda slow
|
Mon, 16 Jun 2008 16:06:59 -0400 |
Yoshua Bengio |
Added untested RenamedFieldsDataSet
|
Mon, 16 Jun 2008 12:57:32 -0400 |
Frederic Bastien |
added file that was missing in commit 9e96fe8b955c
|
Fri, 13 Jun 2008 15:26:35 -0400 |
Olivier Breuleux |
made ExampleWiseMean Op
|
Fri, 13 Jun 2008 11:14:59 -0400 |
Thierry Bertin-Mahieux |
wrote back _cache
|
Thu, 12 Jun 2008 20:54:49 -0400 |
James Bergstra |
- the cut and paste between file and dir conditions is always a bad thing
|
Thu, 12 Jun 2008 17:12:45 -0400 |
Thierry Bertin-Mahieux |
version works with file and folder, returns null if not in hg. Also a function to get all modules, but seems to get junk at the same time
|
Thu, 12 Jun 2008 12:45:42 -0400 |
Thierry Bertin-Mahieux |
small bugs fixed with NArrayDataSet
|
Thu, 12 Jun 2008 12:35:47 -0400 |
Thierry Bertin-Mahieux |
NArraysDataSet improved, use arrays instead of matrix, also a dictionnary of field indexes
|
Wed, 11 Jun 2008 17:11:37 -0400 |
Thierry Bertin-Mahieux |
bug fixed when initializing NArraysDataSet with 1-d arrays. This now raises an Exception
|
Wed, 11 Jun 2008 17:05:56 -0400 |
Thierry Bertin-Mahieux |
bug fixed when one matrix is an array, a 1-d matrix
|
Wed, 11 Jun 2008 16:59:03 -0400 |
Thierry Bertin-Mahieux |
NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff...
|
Wed, 11 Jun 2008 16:40:47 -0400 |
Thierry Bertin-Mahieux |
doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet
|
Wed, 11 Jun 2008 16:28:09 -0400 |
Thierry Bertin-Mahieux |
getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList
|
Wed, 11 Jun 2008 16:26:41 -0400 |
Thierry Bertin-Mahieux |
test to compare overriden __getitem__ implemented, tested on ArrayDataSet.__getitem__
|
Wed, 11 Jun 2008 13:57:34 -0400 |
Thierry Bertin-Mahieux |
bug fixed concerning the slicing, now ds[0:len(ds) + 1000 : 2] is accepted, same a python list
|
Wed, 11 Jun 2008 13:53:39 -0400 |
Thierry Bertin-Mahieux |
behaviour is now the same as a list in pylearn, so if len(ds) = 10, ds[10] raise an IndexError, same thing for ds[[1,10]], and ds[0:14:1] returns 10 elements
|
Wed, 11 Jun 2008 13:22:00 -0400 |
Thierry Bertin-Mahieux |
still something weird in the getitem test
|
Wed, 11 Jun 2008 12:45:24 -0400 |
Thierry Bertin-Mahieux |
a lot of tests are broken because of the new behaviour of __getitem__ that always returns a LookupList, working on that...
|
Wed, 11 Jun 2008 11:43:54 -0400 |
Thierry Bertin-Mahieux |
dataset.subset implemented
|
Wed, 11 Jun 2008 11:18:14 -0400 |
Thierry Bertin-Mahieux |
get item now returns LookupLists
|
Tue, 10 Jun 2008 20:02:25 -0400 |
James Bergstra |
init of version.py
|
Tue, 10 Jun 2008 18:33:15 -0400 |
James Bergstra |
merge
|
Tue, 10 Jun 2008 13:02:04 -0400 |
James Bergstra |
added src_version cache
|
Tue, 10 Jun 2008 17:42:59 -0400 |
Thierry Bertin-Mahieux |
more verbose
|
Tue, 10 Jun 2008 17:16:49 -0400 |
Thierry Bertin-Mahieux |
while learning using __call__, we can now set the early stopper
|