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

Age	Commit message (Collapse)	Author
2016-02-25	Add VolumetricBatchNormalization	Sam Gross
	The BatchNormalization modules now all extend nn.BatchNormalization and use the same THNN/THCUNN implementation.
2016-02-18	Add THNN conversion for Spatial* modules	Francisco Massa
	Add THNN conversion of SpatialBatchNormalization, SpatialFractionalMaxPooling and SpatialSubSampling Add THNN convertion of SpatialConvolutionLocal, SpatialFullConvolution and SpatialUpSamplingNearest THNN conversion of SpatialMaxUnpooling Remove unfold from generic Add functional conversion of SpatialCrossMapLRN Plus fix in the init.c Fix
2016-02-12	clearState save_mean and save_std in BN	Sergey Zagoruyko

2016-02-09	nn.clearState	Sergey Zagoruyko

2016-02-08	Update SpatialBatchNormalization.lua	Timothy Emerick
	Fix typo and copy-paste mistake
2016-02-04	Fix loading of old SpatialBatchNormalization modules	Sam Gross
	The old-style running_std is actually the E[1/sqrt(var + eps)]. I forgot to subtract out the 'eps' when converting to running_var.
2016-01-22	fix batchnorm reset	Sergey Zagoruyko

2016-01-05	Add C implementation of SpatialBatchNormalization	Sam Gross
	This is primarily to support the fast, memory-efficient CUDA implementation. Some other changes include making weight and bias each individually optional and averaging the variances instead of the inverse standard deviation.
2015-10-19	fix batchnorm reset	Sergey Zagoruyko

2015-09-11	30% Memory saving	Dimitrios Korkinof
	Not initialising those variables saved 30% of GPU memory when not training.
2015-06-03	batchnorm is clonable by adding the running estimates to constructor	soumith
	fixing batchnorm tests
2015-04-21	making running_mean/std shareable in batch normalization	soumith

2015-03-21	adding batch normalization	soumith