10:56 < layback> 4

21:04 < palashahuja> are there tests of lstm with rnn in mlpack ?

21:29 < zoq> palashahuja: Hello, take a look at recurrent_network_test.cpp, there is also an LSTM test case in my mlpack fork which hasn't been merged: https://github.com/zoq/mlpack/blob/ann_test/src/mlpack/tests/ann_layer_test.cpp

21:34 < palash> @zoq, is the training happening from lines 737-774 ?

21:34 < palash> with lstm ?

21:35 < palash> I got the idea, though ..

21:40 < zoq> palash: The GradientLSTMLayerTest test case doesn't really train the network, it's a numerical check of the LSTM gradient function. 'struct GradientFunction' is just an abstraction used for the CheckGradient. That way we don't have to write a new CheckGradient function for a specific model.

21:41 < palash> also, can we measure cross entropy error in neural networks ? are there tools present that do so ?

21:50 < zoq> palash: It looks like I haven't updated the cross entropy function for the neural network code, so right now there is no cross entropy error function.

21:53 < zoq> Btw. any contribution is greatly appreciated :)