sparse_nn_single_inference.hpp File Reference

Implements (non-batched) sparse neural network inference. More...

Go to the source code of this file.

Namespaces
	grb
	The ALP/GraphBLAS namespace.
	grb::algorithms
	The namespace for ALP/GraphBLAS algorithms.

Functions
template<Descriptor descr = descriptors::no_operation, typename IOType , typename WeightType , typename BiasType , class ReluMonoid = Monoid< grb::operators::relu< IOType >, grb::identities::negative_infinity >, class Ring = Semiring< grb::operators::add< IOType >, grb::operators::mul< IOType >, grb::identities::zero, grb::identities::one >>
grb::RC	sparse_nn_single_inference (grb::Vector< IOType > &out, const grb::Vector< IOType > &in, const std::vector< grb::Matrix< WeightType > > &layers, const std::vector< BiasType > &biases, grb::Vector< IOType > &temp, const ReluMonoid &relu=ReluMonoid(), const Ring &ring=Ring())
	Performs an inference step of a single data element through a Sparse Neural Network defined by num_layers sparse weight matrices and num_layers biases. More...
template<Descriptor descr = descriptors::no_operation, typename IOType , typename WeightType , typename BiasType , typename ThresholdType = IOType, class MinMonoid = Monoid< grb::operators::min< IOType >, grb::identities::infinity >, class ReluMonoid = Monoid< grb::operators::relu< IOType >, grb::identities::negative_infinity >, class Ring = Semiring< grb::operators::add< IOType >, grb::operators::mul< IOType >, grb::identities::zero, grb::identities::one >>
grb::RC	sparse_nn_single_inference (grb::Vector< IOType > &out, const grb::Vector< IOType > &in, const std::vector< grb::Matrix< WeightType > > &layers, const std::vector< BiasType > &biases, const ThresholdType threshold, grb::Vector< IOType > &temp, const ReluMonoid &relu=ReluMonoid(), const MinMonoid &min=MinMonoid(), const Ring &ring=Ring())
	Performs an inference step of a single data element through a Sparse Neural Network defined by num_layers sparse weight matrices and num_layers biases. More...

Detailed Description

Implements (non-batched) sparse neural network inference.

Author

Aristeidis Mastoras