nerv - Lua-based toolkit for high-performance deep learning

	Commit message (Collapse)	Author	Age
*	change concept of ParamRepo; provide generalized param update; code ↵	Determinant	2015-06-20
\| \| \| \|	clean-up; #25 #26 #27 #29
*	...	Determinant	2015-06-10
\|
*	add CombinerLayer to support branches in NN; add MSELayer	Determinant	2015-06-10
\|
*	fix #20	Determinant	2015-06-09
\|
*	fix #22	Determinant	2015-06-09
\|
*	...	Determinant	2015-06-08
\|
*	add get_params to all layers	Determinant	2015-06-05
\|
*	add profiling; add ce accurarcy; several other changes	Determinant	2015-06-05
\|
*	demo now works (without random shuffle)	Determinant	2015-06-03
\|
*	fix a bug: input[1] should be input[2] (since Lua arrays are 1-based)	Determinant	2015-06-02
\|
*	add ParamRepo, LayerRepo, DAGLayer	Determinant	2015-06-02
\|
*	add preprocessing layers and change layer constructor interface	Determinant	2015-06-02
\|
*	add expand_frm, rearrange_frm, scale_row	Determinant	2015-06-01
\|
*	rename param file to chunk file (intend to generalize it)	Determinant	2015-05-30
\|
*	add softmax + ce layer; test_dnn_layers produces the same result as TNet	Determinant	2015-05-28
\|
*	should support multiple input/output for layers	Determinant	2015-05-28
\|
*	...	Determinant	2015-05-27
\|
*	add implementation for sigmoid layer (not tested)	Determinant	2015-05-27
\|
*	fix a severe bug in memory management of userdata	Determinant	2015-05-26
\|
*	add affine layer implementation	Determinant	2015-05-26
\|
*	add fill for cumatrix	Determinant	2015-05-26
\|
*	add add_row for cumatrix	Determinant	2015-05-26
\|
*	add layer and param	Determinant	2015-05-26