#include <nnet-linear-transform.h>

Inheritance diagram for LinearTransform:

Collaboration diagram for LinearTransform:

[legend]

Public Member Functions
	LinearTransform (int32 dim_in, int32 dim_out)

	~LinearTransform ()

Component *	Copy () const
	Copy component (deep copy),. More...

ComponentType	GetType () const
	Get Type Identification of the component,. More...

void	InitData (std::istream &is)
	Initialize the content of the component by the 'line' from the prototype,. More...

void	ReadData (std::istream &is, bool binary)
	Reads the component content. More...

void	WriteData (std::ostream &os, bool binary) const
	Writes the component content. More...

int32	NumParams () const
	Number of trainable parameters,. More...

void	GetGradient (VectorBase< BaseFloat > *gradient) const
	Get gradient reshaped as a vector,. More...

void	GetParams (VectorBase< BaseFloat > *params) const
	Get the trainable parameters reshaped as a vector,. More...

void	SetParams (const VectorBase< BaseFloat > &params)
	Set the trainable parameters from, reshaped as a vector,. More...

void	SetLinearity (const MatrixBase< BaseFloat > &l)

std::string	Info () const
	Print some additional info (after <ComponentName> and the dims),. More...

std::string	InfoGradient () const
	Print some additional info about gradient (after <...> and dims),. More...

void	PropagateFnc (const CuMatrixBase< BaseFloat > &in, CuMatrixBase< BaseFloat > *out)
	Abstract interface for propagation/backpropagation. More...

void	BackpropagateFnc (const CuMatrixBase< BaseFloat > &in, const CuMatrixBase< BaseFloat > &out, const CuMatrixBase< BaseFloat > &out_diff, CuMatrixBase< BaseFloat > *in_diff)
	Backward pass transformation (to be implemented by descending class...) More...

void	Update (const CuMatrixBase< BaseFloat > &input, const CuMatrixBase< BaseFloat > &diff)
	Compute gradient and update parameters,. More...

const CuMatrixBase< BaseFloat > &	GetLinearity ()
	Accessors to the component parameters. More...

void	SetLinearity (const CuMatrixBase< BaseFloat > &linearity)

const CuMatrixBase< BaseFloat > &	GetLinearityCorr ()

Public Member Functions inherited from UpdatableComponent
	UpdatableComponent (int32 input_dim, int32 output_dim)

virtual	~UpdatableComponent ()

bool	IsUpdatable () const
	Check if contains trainable parameters,. More...

virtual void	SetTrainOptions (const NnetTrainOptions &opts)
	Set the training options to the component,. More...

const NnetTrainOptions &	GetTrainOptions () const
	Get the training options from the component,. More...

virtual void	SetLearnRateCoef (BaseFloat val)
	Set the learn-rate coefficient,. More...

virtual void	SetBiasLearnRateCoef (BaseFloat val)
	Set the learn-rate coefficient for bias,. More...

Public Member Functions inherited from Component
	Component (int32 input_dim, int32 output_dim)
	Generic interface of a component,. More...

virtual	~Component ()

virtual bool	IsMultistream () const
	Check if component has 'Recurrent' interface (trainable and recurrent),. More...

int32	InputDim () const
	Get the dimension of the input,. More...

int32	OutputDim () const
	Get the dimension of the output,. More...

void	Propagate (const CuMatrixBase< BaseFloat > &in, CuMatrix< BaseFloat > *out)
	Perform forward-pass propagation 'in' -> 'out',. More...

void	Backpropagate (const CuMatrixBase< BaseFloat > &in, const CuMatrixBase< BaseFloat > &out, const CuMatrixBase< BaseFloat > &out_diff, CuMatrix< BaseFloat > *in_diff)
	Perform backward-pass propagation 'out_diff' -> 'in_diff'. More...

void	Write (std::ostream &os, bool binary) const
	Write the component to a stream,. More...

Private Attributes
CuMatrix< BaseFloat >	linearity_

CuMatrix< BaseFloat >	linearity_corr_

Additional Inherited Members
Public Types inherited from Component
enum	ComponentType { kUnknown = 0x0, kUpdatableComponent = 0x0100, kAffineTransform, kLinearTransform, kConvolutionalComponent, kLstmProjected, kBlstmProjected, kRecurrentComponent, kActivationFunction = 0x0200, kSoftmax, kHiddenSoftmax, kBlockSoftmax, kSigmoid, kTanh, kParametricRelu, kDropout, kLengthNormComponent, kTranform = 0x0400, kRbm, kSplice, kCopy, kTranspose, kBlockLinearity, kAddShift, kRescale, kKlHmm = 0x0800, kSentenceAveragingComponent, kSimpleSentenceAveragingComponent, kAveragePoolingComponent, kMaxPoolingComponent, kFramePoolingComponent, kParallelComponent, kMultiBasisComponent }
	Component type identification mechanism,. More...

Static Public Member Functions inherited from Component
static const char *	TypeToMarker (ComponentType t)
	Converts component type to marker,. More...

static ComponentType	MarkerToType (const std::string &s)
	Converts marker to component type (case insensitive),. More...

static Component *	Init (const std::string &conf_line)
	Initialize component from a line in config file,. More...

static Component *	Read (std::istream &is, bool binary)
	Read the component from a stream (static method),. More...

Static Public Attributes inherited from Component
static const struct key_value	kMarkerMap []
	The table with pairs of Component types and markers (defined in nnet-component.cc),. More...

Protected Attributes inherited from UpdatableComponent
NnetTrainOptions	opts_
	Option-class with training hyper-parameters,. More...

BaseFloat	learn_rate_coef_
	Scalar applied to learning rate for weight matrices (to be used in ::Update method),. More...

BaseFloat	bias_learn_rate_coef_
	Scalar applied to learning rate for bias (to be used in ::Update method),. More...

Protected Attributes inherited from Component
int32	input_dim_
	Data members,. More...

int32	output_dim_
	Dimension of the output of the Component,. More...

Detailed Description

Definition at line 33 of file nnet-linear-transform.h.

Constructor & Destructor Documentation

◆ LinearTransform()

LinearTransform	(	int32	dim_in,
		int32	dim_out
	)

inline

Definition at line 35 of file nnet-linear-transform.h.

Referenced by LinearTransform::Copy().

                                               :
     UpdatableComponent(dim_in, dim_out),
     linearity_(dim_out, dim_in),
     linearity_corr_(dim_out, dim_in)
   { }

◆ ~LinearTransform()

~LinearTransform ( )

inline

Definition at line 41 of file nnet-linear-transform.h.

42 { }

Member Function Documentation

◆ BackpropagateFnc()

void BackpropagateFnc	(	const CuMatrixBase< BaseFloat > &	in,
		const CuMatrixBase< BaseFloat > &	out,
		const CuMatrixBase< BaseFloat > &	out_diff,
		CuMatrixBase< BaseFloat > *	in_diff
	)

inlinevirtual

Backward pass transformation (to be implemented by descending class...)

Implements Component.

Definition at line 161 of file nnet-linear-transform.h.

References CuMatrixBase< Real >::AddMatMat(), kaldi::kNoTrans, and LinearTransform::linearity_.

                                                           {
     // multiply error derivative by weights
     in_diff->AddMatMat(1.0, out_diff, kNoTrans, linearity_, kNoTrans, 0.0);
   }

◆ Copy()

Component* Copy ( ) const

inlinevirtual

Copy component (deep copy),.

Implements Component.

Definition at line 44 of file nnet-linear-transform.h.

References LinearTransform::LinearTransform().

44 { return new LinearTransform(*this); }

kaldi::nnet1::LinearTransform::LinearTransform

LinearTransform(int32 dim_in, int32 dim_out)

Definition: nnet-linear-transform.h:35

◆ GetGradient()

void GetGradient ( VectorBase< BaseFloat > * gradient ) const

inlinevirtual

Get gradient reshaped as a vector,.

Implements UpdatableComponent.

Definition at line 123 of file nnet-linear-transform.h.

References VectorBase< Real >::CopyRowsFromMat(), VectorBase< Real >::Dim(), KALDI_ASSERT, LinearTransform::linearity_corr_, and LinearTransform::NumParams().

                                                           {
     KALDI_ASSERT(gradient->Dim() == NumParams());
     gradient->CopyRowsFromMat(linearity_corr_);
   }

◆ GetLinearity()

const CuMatrixBase<BaseFloat>& GetLinearity ( )

inline

Accessors to the component parameters.

Definition at line 194 of file nnet-linear-transform.h.

References LinearTransform::linearity_.

194 { return linearity_; }

kaldi::nnet1::LinearTransform::linearity_

CuMatrix< BaseFloat > linearity_

Definition: nnet-linear-transform.h:205

◆ GetLinearityCorr()

const CuMatrixBase<BaseFloat>& GetLinearityCorr ( )

inline

Definition at line 202 of file nnet-linear-transform.h.

References LinearTransform::linearity_corr_.

202 { return linearity_corr_; }

kaldi::nnet1::LinearTransform::linearity_corr_

CuMatrix< BaseFloat > linearity_corr_

Definition: nnet-linear-transform.h:206

◆ GetParams()

void GetParams ( VectorBase< BaseFloat > * params ) const

inlinevirtual

Get the trainable parameters reshaped as a vector,.

Implements UpdatableComponent.

Definition at line 128 of file nnet-linear-transform.h.

References VectorBase< Real >::CopyRowsFromMat(), VectorBase< Real >::Dim(), KALDI_ASSERT, LinearTransform::linearity_, and LinearTransform::NumParams().

                                                       {
     KALDI_ASSERT(params->Dim() == NumParams());
     params->CopyRowsFromMat(linearity_);
   }

◆ GetType()

ComponentType GetType ( ) const

inlinevirtual

Get Type Identification of the component,.

Implements Component.

Definition at line 45 of file nnet-linear-transform.h.

References Component::kLinearTransform.

45 { return kLinearTransform; }

kaldi::nnet1::Component::kLinearTransform

Definition: nnet-component.h:52

◆ Info()

std::string Info ( ) const

inlinevirtual

Print some additional info (after <ComponentName> and the dims),.

Reimplemented from Component.

Definition at line 144 of file nnet-linear-transform.h.

References UpdatableComponent::learn_rate_coef_, LinearTransform::linearity_, kaldi::nnet1::MomentStatistics(), and kaldi::nnet1::ToString().

                          {
     return std::string("\n  linearity") +
       MomentStatistics(linearity_) +
       ", lr-coef " + ToString(learn_rate_coef_);
   }

◆ InfoGradient()

std::string InfoGradient ( ) const

inlinevirtual

Print some additional info about gradient (after <...> and dims),.

Reimplemented from Component.

Definition at line 149 of file nnet-linear-transform.h.

References UpdatableComponent::learn_rate_coef_, LinearTransform::linearity_corr_, kaldi::nnet1::MomentStatistics(), and kaldi::nnet1::ToString().

                                  {
     return std::string("\n  linearity_grad") +
       MomentStatistics(linearity_corr_) +
       ", lr-coef " + ToString(learn_rate_coef_);
   }

◆ InitData()

void InitData ( std::istream & is )

inlinevirtual

Initialize the content of the component by the 'line' from the prototype,.

Implements UpdatableComponent.

Definition at line 47 of file nnet-linear-transform.h.

References Input::Close(), Component::InputDim(), KALDI_ERR, KALDI_LOG, UpdatableComponent::learn_rate_coef_, LinearTransform::linearity_, Component::OutputDim(), kaldi::nnet1::RandGauss(), kaldi::ReadBasicType(), kaldi::ReadToken(), and Input::Stream().

                                 {
     // define options
     float param_stddev = 0.1;
     std::string read_matrix_file;
     // parse config
     std::string token;
     while (is >> std::ws, !is.eof()) {
       ReadToken(is, false, &token);
        if (token == "<ParamStddev>") ReadBasicType(is, false, &param_stddev);
       else if (token == "<ReadMatrix>") ReadToken(is, false, &read_matrix_file);
       else if (token == "<LearnRateCoef>") ReadBasicType(is, false, &learn_rate_coef_);
       else KALDI_ERR << "Unknown token " << token << ", a typo in config?"
                      << " (ParamStddev|ReadMatrix|LearnRateCoef)";
     }
 
     if (read_matrix_file != "") {  // load from file,
       bool binary;
       Input in(read_matrix_file, &binary);
       linearity_.Read(in.Stream(), binary);
       in.Close();
       // check dims,
       if (OutputDim() != linearity_.NumRows() ||
           InputDim() != linearity_.NumCols()) {
         KALDI_ERR << "Dimensionality mismatch! Expected matrix"
                   << " r=" << OutputDim() << " c=" << InputDim()
                   << ", loaded matrix " << read_matrix_file
                   << " with r=" << linearity_.NumRows()
                   << " c=" << linearity_.NumCols();
       }
       KALDI_LOG << "Loaded <LinearTransform> matrix from file : "
                 << read_matrix_file;
       return;
     }
 
     //
     // Initialize trainable parameters,
     //
     // Gaussian with given std_dev (mean = 0),
     linearity_.Resize(OutputDim(), InputDim());
     RandGauss(0.0, param_stddev, &linearity_);
   }

◆ NumParams()

int32 NumParams ( ) const

inlinevirtual

Number of trainable parameters,.

Implements UpdatableComponent.

Definition at line 119 of file nnet-linear-transform.h.

References LinearTransform::linearity_.

Referenced by LinearTransform::GetGradient(), LinearTransform::GetParams(), and LinearTransform::SetParams().

                           {
     return linearity_.NumRows()*linearity_.NumCols();
   }

◆ PropagateFnc()

void PropagateFnc	(	const CuMatrixBase< BaseFloat > &	in,
		CuMatrixBase< BaseFloat > *	out
	)

inlinevirtual

Abstract interface for propagation/backpropagation.

Forward pass transformation (to be implemented by descending class...)

Implements Component.

Definition at line 155 of file nnet-linear-transform.h.

References CuMatrixBase< Real >::AddMatMat(), kaldi::kNoTrans, kaldi::kTrans, and LinearTransform::linearity_.

                                                   {
     // multiply by weights^t
     out->AddMatMat(1.0, in, kNoTrans, linearity_, kTrans, 0.0);
   }

◆ ReadData()

void ReadData	(	std::istream &	is,
		bool	binary
	)

inlinevirtual

Reads the component content.

Reimplemented from Component.

Definition at line 89 of file nnet-linear-transform.h.

References kaldi::ExpectToken(), Component::input_dim_, KALDI_ASSERT, KALDI_ERR, UpdatableComponent::learn_rate_coef_, LinearTransform::linearity_, Component::output_dim_, kaldi::Peek(), kaldi::PeekToken(), kaldi::ReadBasicType(), and kaldi::ReadToken().

                                              {
     // Read all the '<Tokens>' in arbitrary order,
     while ('<' == Peek(is, binary)) {
       int first_char = PeekToken(is, binary);
       switch (first_char) {
         case 'L': ExpectToken(is, binary, "<LearnRateCoef>");
           ReadBasicType(is, binary, &learn_rate_coef_);
           break;
         default:
           std::string token;
           ReadToken(is, false, &token);
           KALDI_ERR << "Unknown token: " << token;
       }
     }
     // Read the data (data follow the tokens),
 
     // weights
     linearity_.Read(is, binary);
 
     KALDI_ASSERT(linearity_.NumRows() == output_dim_);
     KALDI_ASSERT(linearity_.NumCols() == input_dim_);
   }

◆ SetLinearity() [1/2]

void SetLinearity ( const MatrixBase< BaseFloat > & l )

inline

Definition at line 138 of file nnet-linear-transform.h.

References KALDI_ASSERT, LinearTransform::linearity_, MatrixBase< Real >::NumCols(), and MatrixBase< Real >::NumRows().

Referenced by main().

                                                     {
     KALDI_ASSERT(l.NumCols() == linearity_.NumCols());
     KALDI_ASSERT(l.NumRows() == linearity_.NumRows());
     linearity_.CopyFromMat(l);
   }

◆ SetLinearity() [2/2]

void SetLinearity ( const CuMatrixBase< BaseFloat > & linearity )

inline

Definition at line 196 of file nnet-linear-transform.h.

References KALDI_ASSERT, LinearTransform::linearity_, CuMatrixBase< Real >::NumCols(), and CuMatrixBase< Real >::NumRows().

                                                               {
     KALDI_ASSERT(linearity.NumRows() == linearity_.NumRows());
     KALDI_ASSERT(linearity.NumCols() == linearity_.NumCols());
     linearity_.CopyFromMat(linearity);
   }

◆ SetParams()

void SetParams ( const VectorBase< BaseFloat > & params )

inlinevirtual

Set the trainable parameters from, reshaped as a vector,.

Implements UpdatableComponent.

Definition at line 133 of file nnet-linear-transform.h.

References VectorBase< Real >::Dim(), KALDI_ASSERT, LinearTransform::linearity_, and LinearTransform::NumParams().

                                                       {
     KALDI_ASSERT(params.Dim() == NumParams());
     linearity_.CopyRowsFromVec(params);
   }

◆ Update()

void Update	(	const CuMatrixBase< BaseFloat > &	input,
		const CuMatrixBase< BaseFloat > &	diff
	)

inlinevirtual

Compute gradient and update parameters,.

Implements UpdatableComponent.

Definition at line 170 of file nnet-linear-transform.h.

References kaldi::kNoTrans, kaldi::kTrans, NnetTrainOptions::l1_penalty, NnetTrainOptions::l2_penalty, NnetTrainOptions::learn_rate, UpdatableComponent::learn_rate_coef_, LinearTransform::linearity_, LinearTransform::linearity_corr_, NnetTrainOptions::momentum, CuMatrixBase< Real >::NumRows(), UpdatableComponent::opts_, and kaldi::cu::RegularizeL1().

                                                    {
     // we use following hyperparameters from the option class
     const BaseFloat lr = opts_.learn_rate;
     const BaseFloat mmt = opts_.momentum;
     const BaseFloat l2 = opts_.l2_penalty;
     const BaseFloat l1 = opts_.l1_penalty;
     // we will also need the number of frames in the mini-batch
     const int32 num_frames = input.NumRows();
     // compute gradient (incl. momentum)
     linearity_corr_.AddMatMat(1.0, diff, kTrans, input, kNoTrans, mmt);
     // l2 regularization
     if (l2 != 0.0) {
       linearity_.AddMat(-lr*l2*num_frames, linearity_);
     }
     // l1 regularization
     if (l1 != 0.0) {
       cu::RegularizeL1(&linearity_, &linearity_corr_, lr*l1*num_frames, lr);
     }
     // update
     linearity_.AddMat(-lr*learn_rate_coef_, linearity_corr_);
   }

◆ WriteData()

void WriteData	(	std::ostream &	os,
		bool	binary
	)		const

inlinevirtual

Writes the component content.

Reimplemented from Component.

Definition at line 112 of file nnet-linear-transform.h.

References UpdatableComponent::learn_rate_coef_, LinearTransform::linearity_, kaldi::WriteBasicType(), and kaldi::WriteToken().

                                                     {
     WriteToken(os, binary, "<LearnRateCoef>");
     WriteBasicType(os, binary, learn_rate_coef_);
     if (!binary) os << "\n";
     linearity_.Write(os, binary);
   }

Member Data Documentation

◆ linearity_

CuMatrix<BaseFloat> linearity_

private

Definition at line 205 of file nnet-linear-transform.h.

Referenced by LinearTransform::BackpropagateFnc(), LinearTransform::GetLinearity(), LinearTransform::GetParams(), LinearTransform::Info(), LinearTransform::InitData(), LinearTransform::NumParams(), LinearTransform::PropagateFnc(), LinearTransform::ReadData(), LinearTransform::SetLinearity(), LinearTransform::SetParams(), LinearTransform::Update(), and LinearTransform::WriteData().

◆ linearity_corr_

CuMatrix<BaseFloat> linearity_corr_

private

Definition at line 206 of file nnet-linear-transform.h.

Referenced by LinearTransform::GetGradient(), LinearTransform::GetLinearityCorr(), LinearTransform::InfoGradient(), and LinearTransform::Update().

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

nnet/nnet-linear-transform.h

Public Member Functions

Private Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ LinearTransform()

◆ ~LinearTransform()

Member Function Documentation

◆ BackpropagateFnc()

◆ Copy()

◆ GetGradient()

◆ GetLinearity()

◆ GetLinearityCorr()

◆ GetParams()

◆ GetType()

◆ Info()

◆ InfoGradient()

◆ InitData()

◆ NumParams()

◆ PropagateFnc()

◆ ReadData()

◆ SetLinearity() [1/2]

◆ SetLinearity() [2/2]

◆ SetParams()

◆ Update()

◆ WriteData()

Member Data Documentation

◆ linearity_

◆ linearity_corr_