neural

Neural Network. More...

Collaboration diagram for neural:

Functions
void	fold_unfold (Tensor<> &X, Tensor<> &Y)
	folds a tensor X into tensor Y assuming the lexicographical ordering of elements in both tensors is the same but the order is different More...
int	neural (int n, int m, int d, double sp, World &dw)
	computes a neural network iteration for tensor n*n*m tensor X whose sparsity fraction is sp. Filter W is d*d*m and dense. Y_ij = sum_{k=1}^m sum_{a=1}^d sum_{b=1}^d X_{i+a % n, j+b % n, k} * W_{abk} this algorithm assumes n = 0 (mod d) More...
char *	getCmdOption (char **begin, char **end, const std::string &option)
int	main (int argc, char **argv)

Detailed Description

Neural Network.

Function Documentation

void fold_unfold	(	Tensor<> &	X,
		Tensor<> &	Y
	)

folds a tensor X into tensor Y assuming the lexicographical ordering of elements in both tensors is the same but the order is different

Definition at line 12 of file neural_network.cxx.

References CTF::Tensor< dtype >::get_local_data(), and CTF::Tensor< dtype >::write().

char* getCmdOption	(	char **	begin,
		char **	end,
		const std::string &	option
	)

Definition at line 131 of file neural_network.cxx.

Referenced by main().

int main	(	int	argc,
		char **	argv
	)

Definition at line 142 of file neural_network.cxx.

References getCmdOption(), neural(), ctf.core::np(), and ctf.core::rank().

int neural	(	int	n,
		int	m,
		int	d,
		double	sp,
		World &	dw
	)

computes a neural network iteration for tensor n*n*m tensor X whose sparsity fraction is sp. Filter W is d*d*m and dense. Y_ij = sum_{k=1}^m sum_{a=1}^d sum_{b=1}^d X_{i+a % n, j+b % n, k} * W_{abk} this algorithm assumes n = 0 (mod d)

Definition at line 27 of file neural_network.cxx.

References ctf.core::a, ctf.core::b, CTF::Tensor< dtype >::fill_random(), CTF::Tensor< dtype >::fill_sp_random(), CTF::fold_unfold(), CTF::Tensor< dtype >::norm2(), CTF::World::rank, and CTF::Tensor< dtype >::write().

Referenced by main().