cpu_mlp::Network Class Reference

Simple feedforward network with contiguous parameter storage. More...

#include <network.hpp>

Public Member Functions
	Network ()=default
size_t	getSize () const
	Total number of parameters. More...
template<int In, int Out, typename Activation = Linear>
void	addLayer ()
	Append a dense layer to the network. More...
void	bindParams (unsigned int seed=kDefaultSeed)
	Bind and initialize parameters and gradient buffers. More...
const Eigen::MatrixXd &	forward (const Eigen::MatrixXd &input)
	Run a forward pass for a batch of inputs. More...
void	backward (const Eigen::MatrixXd &loss_grad)
	Run a backward pass from output loss gradients. More...
void	zeroGrads ()
	Zero the gradient buffer. More...
double *	getParamsData ()
	Access raw parameter buffer. More...
double *	getGradsData ()
	Access raw gradient buffer. More...
void	setParams (const Eigen::VectorXd &new_params)
	Replace parameters with a new vector. More...
void	getGrads (Eigen::VectorXd &out_grads)
	Copy gradients to output vector. More...
void	test (const Eigen::MatrixXd &inputs, const Eigen::MatrixXd &targets, std::string label="Test Results")
	Evaluate accuracy and MSE for a dataset. More...

Detailed Description

Simple feedforward network with contiguous parameter storage.

Constructor & Destructor Documentation

Network()

cpu_mlp::Network::Network ( )

default

Member Function Documentation

addLayer()

template<int In, int Out, typename Activation = Linear>

void cpu_mlp::Network::addLayer ( )

inline

Append a dense layer to the network.

backward()

void cpu_mlp::Network::backward ( const Eigen::MatrixXd &  loss_grad )

inline

Run a backward pass from output loss gradients.

bindParams()

void cpu_mlp::Network::bindParams ( unsigned int  seed = kDefaultSeed )

inline

Bind and initialize parameters and gradient buffers.

forward()

const Eigen::MatrixXd& cpu_mlp::Network::forward ( const Eigen::MatrixXd &  input )

inline

Run a forward pass for a batch of inputs.

Here is the caller graph for this function:

getGrads()

void cpu_mlp::Network::getGrads ( Eigen::VectorXd &  out_grads )

inline

Copy gradients to output vector.

getGradsData()

double* cpu_mlp::Network::getGradsData ( )

inline

Access raw gradient buffer.

getParamsData()

double* cpu_mlp::Network::getParamsData ( )

inline

Access raw parameter buffer.

getSize()

size_t cpu_mlp::Network::getSize ( ) const

inline

Total number of parameters.

Here is the caller graph for this function:

setParams()

void cpu_mlp::Network::setParams ( const Eigen::VectorXd &  new_params )

inline

Replace parameters with a new vector.

test()

void cpu_mlp::Network::test ( const Eigen::MatrixXd &  inputs, const Eigen::MatrixXd &  targets, std::string  label = "Test Results"  )

inline

Evaluate accuracy and MSE for a dataset.

Here is the call graph for this function:

zeroGrads()

void cpu_mlp::Network::zeroGrads ( )

inline

Zero the gradient buffer.

---

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

• src/network.hpp