geopm::DenseLayer Class Reference

Class to store dense layers and perform operations on the layers' 1D and 2D Tensors, aka vectors and matrices, suitable for use in feed-forward neural networks. More...

#include <DenseLayer.hpp>

Inheritance diagram for geopm::DenseLayer:

Public Member Functions
virtual	~DenseLayer ()=default
virtual TensorOneD	forward (const TensorOneD &input) const =0
	Perform inference using the instance weights and biases. More...
virtual size_t	get_input_dim () const =0
	Get the dimension required for the input TensorOneD. More...
virtual size_t	get_output_dim () const =0
	Get the dimension of the resulting TensorOneD. More...
TensorOneD	operator() (const TensorOneD &input) const

Static Public Member Functions
static std::unique_ptr< DenseLayer >	make_unique (const TensorTwoD &weights, const TensorOneD &biases)
	Returns a unique_ptr to a concrete object constructed using the underlying implementation from an pair of weights and biases. More...

Detailed Description

Class to store dense layers and perform operations on the layers' 1D and 2D Tensors, aka vectors and matrices, suitable for use in feed-forward neural networks.

Constructor & Destructor Documentation

~DenseLayer()

virtual geopm::DenseLayer::~DenseLayer ( )

virtualdefault

Member Function Documentation

forward()

virtual TensorOneD geopm::DenseLayer::forward ( const TensorOneD &  input ) const

pure virtual

Perform inference using the instance weights and biases.

Parameters

[in]

input

TensorOneD vector of input signals.

Exceptions

geopm::Exception

if input dimension is incompatible with the DenseLayer.

Returns

Returns a TensorOneD vector of output values.

Implemented in geopm::DenseLayerImp.

get_input_dim()

virtual size_t geopm::DenseLayer::get_input_dim ( ) const

pure virtual

Get the dimension required for the input TensorOneD.

Returns

Returns a size_t equal to the number of columns of weights

Implemented in geopm::DenseLayerImp.

get_output_dim()

virtual size_t geopm::DenseLayer::get_output_dim ( ) const

pure virtual

Get the dimension of the resulting TensorOneD.

Returns

Returns a size_t equal to the number of rows of weights

Implemented in geopm::DenseLayerImp.

make_unique()

std::unique_ptr< DenseLayer > geopm::DenseLayer::make_unique ( const TensorTwoD &  weights, const TensorOneD &  biases  )

static

Returns a unique_ptr to a concrete object constructed using the underlying implementation from an pair of weights and biases.

Parameters

[in]	weight	matrix (TensorTwoD instance)
[in]	bias	vector (TensorOneD instance)

Exceptions

geopm::Exception	if weights is empty
geopm::Exception	if weights and biases dimensions are incompatible. i.e. if weights rows is inequal to bias dimension.

operator()()

TensorOneD geopm::DenseLayer::operator()

(

const TensorOneD &

input

)

const

---

The documentation for this class was generated from the following files:

• DenseLayer.hpp
• DenseLayer.cpp