dl::layers::Linear< T > Class Template Reference

Linear (fully connected) layer: y = xW^T + b. More...

#include <layers.hpp>

Inheritance diagram for dl::layers::Linear< T >:

Collaboration diagram for dl::layers::Linear< T >:

Public Member Functions
	Linear (size_t in_features, size_t out_features, bool bias=true)
	Constructor.
Variable< T >	forward (const Variable< T > &input) override
	Forward pass: y = xW^T + b.
std::vector< Variable< T > * >	parameters () override
	Get parameters (weight and bias)
Variable< T > &	weight ()
Variable< T > &	bias ()
const Variable< T > &	weight () const
const Variable< T > &	bias () const
Public Member Functions inherited from dl::layers::Module< T >
virtual	~Module ()=default
virtual void	zero_grad ()
	Zero gradients of all parameters.
virtual void	train (bool training=true)
	Set training mode.
virtual void	eval ()
	Set evaluation mode.
bool	is_training () const
	Check if module is in training mode.

Additional Inherited Members
Protected Attributes inherited from dl::layers::Module< T >
bool	training_ = true

Detailed Description

template<typename T>
class dl::layers::Linear< T >

Linear (fully connected) layer: y = xW^T + b.

Definition at line 82 of file layers.hpp.

Constructor & Destructor Documentation

Linear()

template<typename T >

dl::layers::Linear< T >::Linear	(	size_t	in_features,
		size_t	out_features,
		bool	bias = true
	)

Constructor.

Parameters

in_features	Number of input features
out_features	Number of output features
bias	Whether to include bias term

Definition at line 13 of file layers.cpp.

Member Function Documentation

bias() [1/2]

template<typename T >

Variable< T > & dl::layers::Linear< T >::bias ( )

inline

Definition at line 106 of file layers.hpp.

bias() [2/2]

template<typename T >

const Variable< T > & dl::layers::Linear< T >::bias ( ) const

inline

Definition at line 108 of file layers.hpp.

forward()

template<typename T >

Variable< T > dl::layers::Linear< T >::forward ( const Variable< T > &  input )

overridevirtual

Forward pass: y = xW^T + b.

Parameters

input

Input variable of shape (batch_size, in_features)

Returns

Output variable of shape (batch_size, out_features)

Implements dl::layers::Module< T >.

Definition at line 38 of file layers.cpp.

parameters()

template<typename T >

std::vector< Variable< T > * > dl::layers::Linear< T >::parameters ( )

overridevirtual

Get parameters (weight and bias)

Implements dl::layers::Module< T >.

Definition at line 54 of file layers.cpp.

weight() [1/2]

template<typename T >

Variable< T > & dl::layers::Linear< T >::weight ( )

inline

Definition at line 105 of file layers.hpp.

weight() [2/2]

template<typename T >

const Variable< T > & dl::layers::Linear< T >::weight ( ) const

inline

Definition at line 107 of file layers.hpp.

---

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

• include/neural_network/layers.hpp
• src/neural_network/layers.cpp