github.com/torch/nn.git - Unnamed repository; edit this file 'description' to name the repository.

Age	Commit message (Collapse)	Author
2014-02-01	Add extra tests for SparseLinear and fix bug where scale was not being ↵	Páidí Creed
	multiplied into bias updates
2014-01-28	Merge remote-tracking branch 'upstream/master'	Páidí Creed

2013-11-27	less strict about ansi libraries, less warnings	Leon Bottou

2013-11-18	fixed PairwiseDistance type() function bug. Overloaded type now returns self.	Jonathan Tompson

2013-11-18	Fixed a bug in PairwiseDistance where the gradInput table isn't converted ↵	Jonathan Tompson
	when Module.type function is called (this bug has always existed and is not due to the recent changes).
2013-11-06	Merge remote-tracking branch 'upstream/master'	Páidí Creed
	Conflicts: extra/nn/test/test.lua
2013-10-29	Fix SparseLinear	Páidí Creed
	Fixed an issue with SparseLinear and added a test along with a new class SparseJacobian (for testing sparse modules)
2013-10-21	going back on Clement's suggestion. It was a good idea, but we're ↵	Jonathan Tompson
	needlessly loosing performance in the 1D case with the input clone.
2013-10-20	fixed a bug in Pairwise distance when the output Lp norm is zero (which ↵	Jonathan Tompson
	results in a divide by zero issue). Rewrote PairwiseDistance following Clement's suggestion to only have one codepath. Fixed a small bug in extra/test/test.lua where the input to the non-batch fprop test was zero.
2013-10-19	it seems like torch.abs() doesn't have a cuda implementation, so the ↵	Jonathan Tompson
	previous commit would fail when PairwiseDistance():cuda() was called.
2013-10-19	Fixed the bprop in PairwiseDistance for pnorms other than one. The ↵	Jonathan Tompson
	gradInput has always been wrong it seems; the sign of the gradient is correct but the magnitude was wrong. I also added a test in extra/nn/test/test.lua for nn.PairwiseDifference, which tests both the batch and non-batch code paths for a few different p-norms.
2013-10-18	Fixed PairwiseDistance for odd Lp norms	Jonathan Tompson

2013-10-17	init self.output with a single element tensor, not an empty tensor	koray kavukcuoglu

2013-10-17	correct the stochastic case check	koray kavukcuoglu

2013-10-16	Merge pull request #163 from vladmnih/batch	Clement Farabet
	Added minibatch support to MarginRankingCriterion and PairwiseDistance.
2013-10-16	C89	Ronan Collobert

2013-10-16	added strict c89 flags	Ronan Collobert

2013-10-15	add 3D convolution documentation	koray kavukcuoglu

2013-10-15	add 3D max pooling	koray kavukcuoglu

2013-10-13	move index declaragtion into pragma	koray kavukcuoglu

2013-10-13	move k into thread private loop	koray kavukcuoglu

2013-09-27	remove WeightedMSECriterion test since it fails with an unknown module	koray kavukcuoglu

2013-10-08	pkg/nn: C89	Ronan Collobert

2013-09-04	Added support for the partialSum parameter to the SpatialConvolutionCUDA module.	Volodymyr Mnih

2013-09-04	Added minibatch support to MarginRankingCriterion and PairwiseDistance.	Volodymyr Mnih

2013-08-11	luaopen_xxx functions need LUA_EXTERNC	Leon Bottou

2013-06-11	Using a default scale in Module.backward().	Clement Farabet
	This should not affect anything, as modules are always used within containers. Mostly important during testing.
2013-06-11	More CUDA testing.	Clement Farabet

2013-06-08	Exposed padding parameter (SpatialConvolutionCUDA)	Clement Farabet

2013-04-16	CMul should resize the gradInput into the same shape as input.	koray kavukcuoglu

2013-03-26	Fixed a missing </file>.	Ivo Danihelka

2013-03-23	Merge pull request #117 from akfidjeland/linear_nan	koray kavukcuoglu
	Linear:updateGradInput avoids NaN and inf
2013-03-23	Sped up getParameters() in simple situations.	Clement Farabet

2013-03-23	Correct implementation of SpatialMaxPoolingCUDA (Sixin's port)	Clement Farabet

2013-03-22	Merge branch 'master' of github.com:andresy/torch	Clement Farabet

2013-03-22	Added Bias to SpatialConvolutionCUDA.	Clement Farabet

2013-03-20	Added Max Pooling for CUDA.	Clement Farabet

2013-03-20	Integrated CUDA conv routines from cuda-convnet library.	Clement Farabet

2013-03-19	CMulTable updateGradInput now uses proper but slow gradient calculation. The ↵	koray kavukcuoglu
	old one is available in updateGradInput_efficient function.
2013-03-13	Linear:updateGradInput avoids NaN and inf	Andreas Fidjeland
	The resize in Linear:updateGradInput can introduce NaN and inf into the gradients. The resize itself leaves garbage in the gradInput tensor. For normal numbers the subsequent addmm/addmv will clear the garbage. However, if gradInput contains either nan or inf after the resize, the multiply will result in nan instead of the desired result.
2013-02-23	Merge pull request #115 from fidlej/topic_Copy	Clement Farabet
	Fixed nn.Copy with default arguments.
2013-02-23	Fixed nn.Copy with default arguments.	Ivo Danihelka

2013-02-23	Corrected getParameters() documentation.	Ivo Danihelka

2013-02-20	Merge branch 'nn_fast_reset'	Clement Farabet

2013-02-20	Merge branch 'master' into nn_fast_reset	Clement Farabet

2013-01-03	New NN classes	koray kavukcuoglu
	extra/nn/L1Cost.lua : L1 penalty extra/nn/SpatialFullConvolution.lua : full convolution extra/nn/SpatialFullConvolutionMap.lua : full convolution with connection table extra/nn/TanhShrink.lua : shrinkage with x-tanh(x) extra/nn/WeightedMSECriterion.lua : mean squared error with weighting mask on the target Add new nn classes that are used commonly for unsupervised training of convolutional auto encoders
2012-12-07	Document function getParameters()	Julien

2012-12-06	fix Jacobian when using direct-updates	Ronan Collobert

2012-11-30	Updated flatten() in hessian.lua to be synchronized the version in ↵	Ivo Danihelka
	Module:getParameters().
2012-11-29	Improved Module:getParameters() speed when using many storages.	Ivo Danihelka