mlask::FullyConnectedLayer< in, out > Class Template Reference

Class representing fully connected layer. More...

#include <FullyConnectedLayer.hpp>

Inheritance diagram for mlask::FullyConnectedLayer< in, out >:

Collaboration diagram for mlask::FullyConnectedLayer< in, out >:

Public Member Functions
	FullyConnectedLayer ()
	A constructor for fully connected layer, initializes weights and biases with random values, and changes with zeros.
vectorOut	forward (vectorIn input) override
	A function defining moving foraward in neural network.
vectorIn	backward (vectorOut error) override
	A backtrack for backpropagation algorithm.
void	fit (float_t learning_rate) override
	A function that updates weights and biases.
bool	tryConvertToONNX (onnx::GraphProto *graph, std::string input, std::string output) const override
	A function that tries to convert the layer to ONNX format, returns true if successful, false otherwise.
std::string	str () const override
	Returns a string representation of the layer.
Eigen::Matrix< float_t, out, in >	weights ()
Eigen::Matrix< float_t, in, 1 >	bias ()
Public Member Functions inherited from mlask::Layer
std::size_t	getIn () const
std::size_t	getOut () const
virtual	~Layer ()=default

Additional Inherited Members
Protected Attributes inherited from mlask::Layer
std::size_t	in_
std::size_t	out_

Detailed Description

template<std::size_t in, std::size_t out>
class mlask::FullyConnectedLayer< in, out >

Class representing fully connected layer.

Meaning a layer with in number of neurons as an input and out number of neurons as an output,

where all the neurons are connected with each other.

Template Parameters

in	The number of neurons in the input
out	The number of neurons in the output

Constructor & Destructor Documentation

FullyConnectedLayer()

template<std::size_t in, std::size_t out>

mlask::FullyConnectedLayer< in, out >::FullyConnectedLayer ( )

inline

A constructor for fully connected layer, initializes weights and biases with random values, and changes with zeros.

Member Function Documentation

backward()

template<std::size_t in, std::size_t out>

vectorIn mlask::FullyConnectedLayer< in, out >::backward ( vectorOut  error )

overridevirtual

A backtrack for backpropagation algorithm.

Implements mlask::Layer.

bias()

template<std::size_t in, std::size_t out>

Eigen::Matrix< float_t, in, 1 > mlask::FullyConnectedLayer< in, out >::bias ( )

inline

fit()

template<std::size_t in, std::size_t out>

void mlask::FullyConnectedLayer< in, out >::fit ( float_t  learning_rate )

overridevirtual

A function that updates weights and biases.

Implements mlask::Layer.

forward()

template<std::size_t in, std::size_t out>

vectorOut mlask::FullyConnectedLayer< in, out >::forward ( vectorIn  input )

overridevirtual

A function defining moving foraward in neural network.

Implements mlask::Layer.

str()

template<std::size_t in, std::size_t out>

std::string mlask::FullyConnectedLayer< in, out >::str ( ) const

overridevirtual

Returns a string representation of the layer.

Reimplemented from mlask::Layer.

tryConvertToONNX()

template<std::size_t in, std::size_t out>

bool mlask::FullyConnectedLayer< in, out >::tryConvertToONNX ( onnx::GraphProto *  graph, std::string  input, std::string  output  ) const

overridevirtual

A function that tries to convert the layer to ONNX format, returns true if successful, false otherwise.

Parameters

graph	A pointer to the ONNX graph to which the layer should be added
input	The name of the input tensor for this layer in the ONNX graph
output	The name of the output tensor for this layer in the ONNX graph

Reimplemented from mlask::Layer.

weights()

template<std::size_t in, std::size_t out>

Eigen::Matrix< float_t, out, in > mlask::FullyConnectedLayer< in, out >::weights ( )

inline

---

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

• include/layers/FullyConnectedLayer.hpp