Configuration class that controls neural net "widening", which means increasing the dimension of the hidden layers of an already-trained neural net. More...

#include <widen-nnet.h>

Collaboration diagram for NnetWidenConfig:

[legend]

Public Member Functions
	NnetWidenConfig ()

void	Register (OptionsItf *opts)

Public Attributes
int32	hidden_layer_dim

BaseFloat	param_stddev_factor

BaseFloat	bias_stddev

Detailed Description

Configuration class that controls neural net "widening", which means increasing the dimension of the hidden layers of an already-trained neural net.

Definition at line 33 of file widen-nnet.h.

Constructor & Destructor Documentation

◆ NnetWidenConfig()

NnetWidenConfig ( )

inline

Definition at line 38 of file widen-nnet.h.

                    : hidden_layer_dim(-1),
                      param_stddev_factor(1.0),
                      bias_stddev(0.5) { }

Member Function Documentation

◆ Register()

void Register ( OptionsItf * opts )

inline

Definition at line 42 of file widen-nnet.h.

References OptionsItf::Register(), and kaldi::nnet2::WidenNnet().

Referenced by main().

                                   {
     opts->Register("hidden-layer-dim", &hidden_layer_dim, "[required option]: "
                    "target dimension of hidden layers");
     opts->Register("param-stddev-factor", &param_stddev_factor, "Factor in "
                    "standard deviation of linear parameters of new part of "
                    "transform (multiply by 1/sqrt of input-dim)");
     opts->Register("bias-stddev", &bias_stddev, "Standard deviation of added "
                    "bias parameters");
   }  