Configuration class that controls neural net combination, where we combine a number of neural nets, trying to find for each layer the optimal weighted combination of the different neural-net parameters. More...

#include <combine-nnet.h>

Collaboration diagram for NnetCombineConfig:

[legend]

Public Member Functions
	NnetCombineConfig ()

void	Register (OptionsItf *opts)

Public Attributes
int32	initial_model

int32	num_bfgs_iters

BaseFloat	initial_impr

bool	test_gradient

Detailed Description

Configuration class that controls neural net combination, where we combine a number of neural nets, trying to find for each layer the optimal weighted combination of the different neural-net parameters.

Definition at line 35 of file combine-nnet.h.

Constructor & Destructor Documentation

◆ NnetCombineConfig()

NnetCombineConfig ( )

inline

Definition at line 46 of file combine-nnet.h.

                      : initial_model(-1), num_bfgs_iters(30),
                        initial_impr(0.01),
                        test_gradient(false) { }

Member Function Documentation

◆ Register()

void Register ( OptionsItf * opts )

inline

Definition at line 50 of file combine-nnet.h.

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

Referenced by main().

                                   {
     opts->Register("initial-model", &initial_model, "Specifies where to start the "
                    "optimization from.  If 0 ... #models-1, then specifies the model; "
                    "if #models, then the average of all inputs; otherwise, chosen "
                    "automatically from the previous options.");
     opts->Register("num-bfgs-iters", &num_bfgs_iters, "Maximum number of function "
                    "evaluations for BFGS to use when optimizing combination weights");
     opts->Register("initial-impr", &initial_impr, "Amount of objective-function change "
                    "we aim for on the first iteration.");
     opts->Register("test-gradient", &test_gradient, "If true, activate code that "
                    "tests the gradient is accurate.");
   }  