Convolutional Predictive Coding layer (im2col-based). See the file-level warning above before enabling an unclamped middle layer. More...

#include <ConvPCLayer.h>

Inheritance diagram for Deep::ConvPCLayer:

Public Member Functions
	ConvPCLayer (int inChannels, int outChannels, int inHeight, int inWidth, int kernelH, int kernelW, int strideH=1, int strideW=1, int padH=0, int padW=0, int batchSize=1, float learningRate=1e-6f, float inferenceRate=0.1f, float precisionRate=0.01f, float lmbda=1e-2f, ActivationType aType=ActivationType::RELU, ActivationType dType=ActivationType::dRELU)
	Constructor for a convolutional PC layer.
float	CalculateState () noexcept override
	Calculate energy/prediction errors for this layer.
void	UpdateState () noexcept override
	Update latent beliefs (z/r) via inference gradient.
void	UpdateWeights () noexcept override
	Hebbian/gradient weight update.
void	UpdatePrecision () noexcept
	Updates this layer's precision estimate from current prediction errors, using the configured precision rate (pr).
void	Flush () noexcept override
	No-op; exists for Layer interface conformance.
void	ClampState (const std::vector< float > &inputData) noexcept
	Clamps this layer's beliefs to externally-provided data, preventing them from being updated by UpdateState().
void	UnclampState () noexcept
	Releases a previous ClampState() call, allowing this layer's beliefs to update normally again.
float *	GetBeliefs () noexcept override
	Returns this layer's belief buffer.
const float *	GetErrors () const noexcept override
	Returns this layer's prediction-error buffer.
const float *	GetMu () const noexcept
	Returns this layer's top-down prediction buffer.
size_t	GetInputSize () const noexcept override
	Flattened element count per batch item (inChannelsHW).
size_t	GetOutputSize () const noexcept override
	Flattened element count of this layer's OUTGOING prediction (outChannelsoutHoutW), 0 for a terminal layer.
size_t	GetBatchSize () const noexcept override
	Returns the batch size this layer was constructed with.
const float *	GetWeights () const noexcept
	Returns this layer's weight buffer.
float *	GetWeights () noexcept
	Returns this layer's weight buffer.
const float *	GetBiases () const noexcept
	Returns this layer's bias buffer.
float *	GetBiases () noexcept
	Returns this layer's bias buffer.
const float *	GetPrecisions () const noexcept
	Returns this layer's precision buffer.
float	GetLearningRate () const noexcept
	Returns the learning rate used for weight updates.
float	GetInferenceRate () const noexcept
	Returns the learning rate used for internal-state updates.
float	GetPrecisionRate () const noexcept
	Returns the learning rate used for precision updates.
float	GetLambda () const noexcept
	Returns the weight-decay (L2 regularization) coefficient.
void	SetLearningRate (float lr) noexcept
	Sets the learning rate used for weight updates.
void	SetInferenceRate (float ir) noexcept
	Sets the learning rate used for internal-state updates.
void	SetPrecisionRate (float pr) noexcept
	Sets the learning rate used for precision updates.
void	SetLambda (float l) noexcept
	Sets the weight-decay (L2 regularization) coefficient.
void	SetLayerAbove (ConvPCLayer *above) noexcept
	Sets the layer immediately above this one in the network.
void	SetLayerBelow (ConvPCLayer *below) noexcept
	Sets the layer immediately below this one in the network.
void	ResetState () noexcept
	Resets this layer's beliefs/errors back to their initial values, without touching learned weights.
void	RandomizeWeights (std::mt19937 &twister) noexcept
	Randomizes this layer's weights (and biases) in place.
std::map< std::string, TensorDescriptor >	GetStateDict () const override
	Returns a dictionary of this.
void	ResyncLogPrecision () noexcept
	Rebuilds log_p from p – required after a checkpoint load that only persists p (mirrors DiscriminativePCLayer's fix for the same p/log_p desync issue found in ModelIO::Load()).
void	ComputeMuOnly () noexcept
	Fast-path forward projection that skips state initialization.
ActivationType	GetActivationType () const noexcept
	Returns this layer's configured activation type.
ActivationType	GetDerivativeType () const noexcept
	Returns this layer's configured activation-derivative type.
int	GetInChannels () const noexcept
	Returns the number of input channels.
int	GetOutChannels () const noexcept
	Returns the number of output channels (0 for a terminal layer).
int	GetInHeight () const noexcept
	Returns the input feature-map height.
int	GetInWidth () const noexcept
	Returns the input feature-map width.
int	GetOutHeight () const noexcept
	Returns the output feature-map height.
int	GetOutWidth () const noexcept
	Returns the output feature-map width.
int	GetKernelH () const noexcept
	Returns the convolution kernel height.
int	GetKernelW () const noexcept
	Returns the convolution kernel width.
size_t	GetRequiredFloats () const noexcept
	Computes the total number of floats this layer requires from a MemoryArena (weights, biases, beliefs, errors, and every scratch buffer combined).
void	BindMemory (MemoryArena &arena)
	Binds this layer's weight/state/scratch buffers into the supplied arena. Must be called before any other operation.
Public Member Functions inherited from Deep::Layer
virtual	~Layer ()=default

Friends
class	PCNDiagnostics

Additional Inherited Members
Protected Attributes inherited from Deep::Layer
size_t	size
	Size of input.
size_t	nextSize
	Size of output.

Detailed Description

Convolutional Predictive Coding layer (im2col-based). See the file-level warning above before enabling an unclamped middle layer.

Constructor & Destructor Documentation

◆ ConvPCLayer()

Deep::ConvPCLayer::ConvPCLayer	(	int	inChannels,
		int	outChannels,
		int	inHeight,
		int	inWidth,
		int	kernelH,
		int	kernelW,
		int	strideH = 1,
		int	strideW = 1,
		int	padH = 0,
		int	padW = 0,
		int	batchSize = 1,
		float	learningRate = 1e-6f,
		float	inferenceRate = 0.1f,
		float	precisionRate = 0.01f,
		float	lmbda = 1e-2f,
		ActivationType	aType = ActivationType::RELU,
		ActivationType	dType = ActivationType::dRELU )

Constructor for a convolutional PC layer.

Parameters

inChannels	Number of input channels
outChannels	Number of output channels (0 marks a terminal layer, matching DiscriminativePCLayer's convention where nextSize=0 means "no outgoing prediction")
inHeight	Input feature-map height
inWidth	Input feature-map width
kernelH	Kernel height
kernelW	Kernel width
strideH	Vertical stride
strideW	Horizontal stride
padH	Vertical zero-padding
padW	Horizontal zero-padding
batchSize	Batch size
learningRate	Learning rate for weights
inferenceRate	Learning rate for internal state
precisionRate	Learning rate for precision
lmbda	Weight decay (L2 regularization) coefficient
aType	Activation type
dType	Activation derivative type

Member Function Documentation

◆ BindMemory()

void Deep::ConvPCLayer::BindMemory ( MemoryArena & arena )

Binds this layer's weight/state/scratch buffers into the supplied arena. Must be called before any other operation.

Parameters

arena The MemoryArena to bind into.

◆ CalculateState()

float Deep::ConvPCLayer::CalculateState ( )

overridevirtualnoexcept

Calculate energy/prediction errors for this layer.

Warning: Weight-gradient path verified; feedback term NOT YET verified—see file-level warning.

Returns: This layer's energy contribution at the current state.

Implements Deep::Layer.

◆ ClampState()

void Deep::ConvPCLayer::ClampState ( const std::vector< float > & inputData )

noexcept

Clamps this layer's beliefs to externally-provided data, preventing them from being updated by UpdateState().

Parameters

inputData Flattened input data matching (inChannels, inHeight, inWidth) per batch item.

◆ ComputeMuOnly()

void Deep::ConvPCLayer::ComputeMuOnly ( )

noexcept

Fast-path forward projection that skips state initialization.

◆ Flush()

void Deep::ConvPCLayer::Flush ( )

inlineoverridevirtualnoexcept

No-op; exists for Layer interface conformance.

Reimplemented from Deep::Layer.

◆ GetActivationType()

ActivationType Deep::ConvPCLayer::GetActivationType ( ) const

inlinenoexcept

Returns this layer's configured activation type.

Returns: The activation type.

◆ GetBatchSize()

size_t Deep::ConvPCLayer::GetBatchSize ( ) const

inlineoverridevirtualnoexcept

Returns the batch size this layer was constructed with.

Returns: The batch size.

Implements Deep::Layer.

◆ GetBeliefs()

float * Deep::ConvPCLayer::GetBeliefs ( )

inlineoverridevirtualnoexcept

Returns this layer's belief buffer.

Returns: Pointer to (outChannels, outHeight, outWidth) beliefs per batch item.

Implements Deep::Layer.

◆ GetBiases() [1/2]

const float * Deep::ConvPCLayer::GetBiases ( ) const

inlinenoexcept

Returns this layer's bias buffer.

Returns: Pointer to (outChannels,) biases.

◆ GetBiases() [2/2]

float * Deep::ConvPCLayer::GetBiases ( )

inlinenoexcept

Returns this layer's bias buffer.

Returns: Pointer to (outChannels,) biases.

◆ GetDerivativeType()

ActivationType Deep::ConvPCLayer::GetDerivativeType ( ) const

inlinenoexcept

Returns this layer's configured activation-derivative type.

Returns: The activation-derivative type.

◆ GetErrors()

const float * Deep::ConvPCLayer::GetErrors ( ) const

inlineoverridevirtualnoexcept

Returns this layer's prediction-error buffer.

Returns: Pointer to (outChannels, outHeight, outWidth) errors per batch item.

Implements Deep::Layer.

◆ GetInChannels()

int Deep::ConvPCLayer::GetInChannels ( ) const

inlinenoexcept

Returns the number of input channels.

◆ GetInferenceRate()

float Deep::ConvPCLayer::GetInferenceRate ( ) const

inlinenoexcept

Returns the learning rate used for internal-state updates.

Returns: The inference rate.

◆ GetInHeight()

int Deep::ConvPCLayer::GetInHeight ( ) const

inlinenoexcept

Returns the input feature-map height.

◆ GetInputSize()

size_t Deep::ConvPCLayer::GetInputSize ( ) const

inlineoverridevirtualnoexcept

Flattened element count per batch item (inChannels*H*W).

Implements Deep::Layer.

◆ GetInWidth()

int Deep::ConvPCLayer::GetInWidth ( ) const

inlinenoexcept

Returns the input feature-map width.

◆ GetKernelH()

int Deep::ConvPCLayer::GetKernelH ( ) const

inlinenoexcept

Returns the convolution kernel height.

◆ GetKernelW()

int Deep::ConvPCLayer::GetKernelW ( ) const

inlinenoexcept

Returns the convolution kernel width.

◆ GetLambda()

float Deep::ConvPCLayer::GetLambda ( ) const

inlinenoexcept

Returns the weight-decay (L2 regularization) coefficient.

Returns: Lambda.

◆ GetLearningRate()

float Deep::ConvPCLayer::GetLearningRate ( ) const

inlinenoexcept

Returns the learning rate used for weight updates.

Returns: The learning rate.

◆ GetMu()

const float * Deep::ConvPCLayer::GetMu ( ) const

inlinenoexcept

Returns this layer's top-down prediction buffer.

◆ GetOutChannels()

int Deep::ConvPCLayer::GetOutChannels ( ) const

inlinenoexcept

Returns the number of output channels (0 for a terminal layer).

◆ GetOutHeight()

int Deep::ConvPCLayer::GetOutHeight ( ) const

inlinenoexcept

Returns the output feature-map height.

◆ GetOutputSize()

size_t Deep::ConvPCLayer::GetOutputSize ( ) const

inlineoverridevirtualnoexcept

Flattened element count of this layer's OUTGOING prediction (outChannels*outH*outW), 0 for a terminal layer.

Implements Deep::Layer.

◆ GetOutWidth()

int Deep::ConvPCLayer::GetOutWidth ( ) const

inlinenoexcept

Returns the output feature-map width.

◆ GetPrecisionRate()

float Deep::ConvPCLayer::GetPrecisionRate ( ) const

inlinenoexcept

Returns the learning rate used for precision updates.

Returns: The precision rate.

◆ GetPrecisions()

const float * Deep::ConvPCLayer::GetPrecisions ( ) const

inlinenoexcept

Returns this layer's precision buffer.

Returns: Pointer to (outChannels,) precisions.

◆ GetRequiredFloats()

size_t Deep::ConvPCLayer::GetRequiredFloats ( ) const

noexcept

Computes the total number of floats this layer requires from a MemoryArena (weights, biases, beliefs, errors, and every scratch buffer combined).

Returns: The required float count.

◆ GetStateDict()

std::map< std::string, TensorDescriptor > Deep::ConvPCLayer::GetStateDict ( ) const

overridevirtual

Returns a dictionary of this.

Reimplemented from Deep::Layer.

◆ GetWeights() [1/2]

const float * Deep::ConvPCLayer::GetWeights ( ) const

inlinenoexcept

Returns this layer's weight buffer.

Returns: Pointer to (outChannels, inChannels*kernelH*kernelW) weights.

◆ GetWeights() [2/2]

float * Deep::ConvPCLayer::GetWeights ( )

inlinenoexcept

Returns this layer's weight buffer.

Returns: Pointer to (outChannels, inChannels*kernelH*kernelW) weights.

◆ RandomizeWeights()

void Deep::ConvPCLayer::RandomizeWeights ( std::mt19937 & twister )

noexcept

Randomizes this layer's weights (and biases) in place.

Parameters

twister The classic Mersenne Twister

◆ ResetState()

void Deep::ConvPCLayer::ResetState ( )

noexcept

Resets this layer's beliefs/errors back to their initial values, without touching learned weights.

◆ ResyncLogPrecision()

void Deep::ConvPCLayer::ResyncLogPrecision ( )

noexcept

Rebuilds log_p from p – required after a checkpoint load that only persists p (mirrors DiscriminativePCLayer's fix for the same p/log_p desync issue found in ModelIO::Load()).

◆ SetInferenceRate()

void Deep::ConvPCLayer::SetInferenceRate ( float ir )

inlinenoexcept

Sets the learning rate used for internal-state updates.

Parameters

ir	The new inference rate.

◆ SetLambda()

void Deep::ConvPCLayer::SetLambda ( float l )

inlinenoexcept

Sets the weight-decay (L2 regularization) coefficient.

Parameters

l	The new lambda value.

◆ SetLayerAbove()

void Deep::ConvPCLayer::SetLayerAbove ( ConvPCLayer * above )

inlinenoexcept

Sets the layer immediately above this one in the network.

Parameters

above Pointer to the layer above; may be nullptr for a terminal layer.

◆ SetLayerBelow()

void Deep::ConvPCLayer::SetLayerBelow ( ConvPCLayer * below )

inlinenoexcept

Sets the layer immediately below this one in the network.

Parameters

below Pointer to the layer below; may be nullptr for the input layer.

◆ SetLearningRate()

void Deep::ConvPCLayer::SetLearningRate ( float lr )

inlinenoexcept

Sets the learning rate used for weight updates.

Parameters

lr	The new learning rate.

◆ SetPrecisionRate()

void Deep::ConvPCLayer::SetPrecisionRate ( float pr )

inlinenoexcept

Sets the learning rate used for precision updates.

Parameters

pr	The new precision rate.

◆ UnclampState()

void Deep::ConvPCLayer::UnclampState ( )

noexcept

Releases a previous ClampState() call, allowing this layer's beliefs to update normally again.

◆ UpdatePrecision()

void Deep::ConvPCLayer::UpdatePrecision ( )

noexcept

Updates this layer's precision estimate from current prediction errors, using the configured precision rate (pr).

◆ UpdateState()

void Deep::ConvPCLayer::UpdateState ( )

overridevirtualnoexcept

Update latent beliefs (z/r) via inference gradient.

Warning: Feedback term not yet verified—see file-level warning.

Implements Deep::Layer.

◆ UpdateWeights()

void Deep::ConvPCLayer::UpdateWeights ( )

overridevirtualnoexcept

Hebbian/gradient weight update.

Warning: Weight-gradient path verified; feedback term NOT YET verified—see file-level warning.

Implements Deep::Layer.

◆ PCNDiagnostics

friend class PCNDiagnostics

friend

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

include/deepity/layers/ConvPCLayer.h
src/layers/ConvPCLayer.cpp

Public Member Functions

Friends

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ ConvPCLayer()

Member Function Documentation

◆ BindMemory()

◆ CalculateState()

◆ ClampState()

◆ ComputeMuOnly()

◆ Flush()

◆ GetActivationType()

◆ GetBatchSize()

◆ GetBeliefs()

◆ GetBiases() [1/2]

◆ GetBiases() [2/2]

◆ GetDerivativeType()

◆ GetErrors()

◆ GetInChannels()

◆ GetInferenceRate()

◆ GetInHeight()

◆ GetInputSize()

◆ GetInWidth()

◆ GetKernelH()

◆ GetKernelW()

◆ GetLambda()

◆ GetLearningRate()

◆ GetMu()

◆ GetOutChannels()

◆ GetOutHeight()

◆ GetOutputSize()

◆ GetOutWidth()

◆ GetPrecisionRate()

◆ GetPrecisions()

◆ GetRequiredFloats()

◆ GetStateDict()

◆ GetWeights() [1/2]

◆ GetWeights() [2/2]

◆ RandomizeWeights()

◆ ResetState()

◆ ResyncLogPrecision()

◆ SetInferenceRate()

◆ SetLambda()

◆ SetLayerAbove()

◆ SetLayerBelow()

◆ SetLearningRate()

◆ SetPrecisionRate()

◆ UnclampState()

◆ UpdatePrecision()

◆ UpdateState()

◆ UpdateWeights()

Friends And Related Symbol Documentation

◆ PCNDiagnostics