#include <DiscriminativePCLayer.h>

Inheritance diagram for Deep::DiscriminativePCLayer:

Public Member Functions
	DiscriminativePCLayer (int size, int nextSize, int batchSize=1, float learningRate=1e-6, float inferenceRate=0.1f, float precisionRate=0.01f, float lmbda=1e-2f, void(act)(float , size_t)=relu, void(dAct)(float , size_t, bool)=dRelu)
	Constructor for a Deepity Layer.
	DiscriminativePCLayer (int size, int nextSize, int batchSize=1, float learningRate=1e-6, float inferenceRate=0.1f, float precisionRate=0.01f, float lmbda=1e-2f, ActivationType aType=ActivationType::RELU, ActivationType dType=ActivationType::dRELU)
	Constructor for a Deepity Layer, using a named ActivationType instead of raw function pointers.
float	CalculateState () noexcept override
	Calculates the total network energy state.
void	UpdateState () noexcept override
	Computes the state derivatives for inference.
void	UpdateWeights () noexcept override
	Computes weight updates via gradient descent, with L2 weight decay.
void	UpdatePrecision () noexcept
	Updates this layer's precision weighting based on the current prediction error.
void	Flush () noexcept override
	Does nothing; exists for class extension.
void	ResyncLogPrecision () noexcept
	Recomputes log of precision if it falls behind.
void	ClampState (const std::vector< float > &inputData) noexcept
	Clamps the layer to the input data.
void	UnclampState () noexcept
	Unclamps the layer to the input data.
float *	GetBeliefs () noexcept override
	Returns beliefs.
const float *	GetErrors () const noexcept override
	Returns errors.
const float *	GetMu () const noexcept
	Returns this layer's outgoing prediction buffer.
size_t	GetInputSize () const noexcept override
	Returns size (input size).
size_t	GetOutputSize () const noexcept override
	Returns nextSize (output size).
size_t	GetBatchSize () const noexcept override
	Returns batchSize.
const float *	GetWeights () const noexcept
	Returns a read-only version of the stored weights.
float *	GetWeights () noexcept
	Returns a mutable version of the stored weights.
const float *	GetBiases () const noexcept
	Returns a read-only version of the stored biases.
float *	GetBiases () noexcept
	Returns a mutable version of the stored biases.
const float *	GetPrecisions () const
	Returns a read-only version of the stored precisions.
float	GetLearningRate () const noexcept
	Returns the learning rate used for weight updates.
float	GetInferenceRate () const noexcept
	Returns the inference rate used for 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 inference rate used for 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	SetOptimizer (OptimizerType o) noexcept
	Selects the optimizer used for weight updates.
void	SetLayerAbove (DiscriminativePCLayer *above) noexcept
	Ties this layer to one above it.
void	SetLayerBelow (DiscriminativePCLayer *below) noexcept
	Ties this layer to one below it.
void	ResetState () noexcept
	Resets 'z' (beliefs) to 0.
const DiscriminativePCLayer &	GetLayerAbove () const noexcept
	Gets const pointer to layer above.
const DiscriminativePCLayer &	GetLayerBelow () const noexcept
	Gets const pointer to layer below.
void	RandomizeWeights (std::mt19937 &twister) noexcept
	Makes the weights W randomized to [0.0, 0.1] using an OpenMP-parallelized uniform real distribution.
ActivationType	GetActivationType () const noexcept
	Returns this layer's configured activation type.
ActivationType	GetDerivativeType () const noexcept
	Returns this layer's configured activation-derivative type.
size_t	GetRequiredFloats () const noexcept
	Calculates exact float count required by this layer's architecture.
void	BindMemory (MemoryArena &arena)
	Binds all tensor pointers to the contiguous memory block.
void	ComputeMuOnly () noexcept
	Computes only mu (forward prediction), skipping error and energy computation entirely. Useful for callers that only need a forward pass through current weights (e.g. forward projection initialization) without the cost of the discarded error/energy values.
Public Member Functions inherited from Deep::Layer
virtual	~Layer ()=default
virtual std::map< std::string, TensorDescriptor >	GetStateDict () const
	Returns a dictionary of this.

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

Represents a single layer in a Predictive Coding Network.

To perform inference (prediction):

Clamp the input data to the bottom layer's latent state (z).
Enter a continuous loop across all layers, calling CalculateState() to compute the local prediction errors (e).
Call UpdateState() to adjust the latent states (z) based on those errors.
Repeat steps 2 and 3 until the states settle into an equilibrium (the energy is minimized and dz/dt approaches zero).
Read the final predictions from the latent states of the desired layers.
(For learning): Call UpdateWeights() on all layers simultaneously once the states have fully settled.

Constructor & Destructor Documentation

◆ DiscriminativePCLayer() [1/2]

Deep::DiscriminativePCLayer::DiscriminativePCLayer	(	int	size,
		int	nextSize,
		int	batchSize = 1,
		float	learningRate = 1e-6,
		float	inferenceRate = 0.1f,
		float	precisionRate = 0.01f,
		float	lmbda = 1e-2f,
		void(*	act )(float *, size_t) = relu,
		void(*	dAct )(float *, size_t, bool) = dRelu )

Constructor for a Deepity Layer.

Parameters

size	Size of layer
nextSize	Size of next layer (used for row size of W)
batchSize	Batch size (default to 1 for simplicity)
learningRate	Learning rate to update weights
inferenceRate	Learning rate to update state
precisionRate	Precition weight to update precision
lmbda	Weight decay (L2 regularization) coefficient
act	Activation function
dAct	Derivative of Activation function

◆ DiscriminativePCLayer() [2/2]

Deep::DiscriminativePCLayer::DiscriminativePCLayer	(	int	size,
		int	nextSize,
		int	batchSize = 1,
		float	learningRate = 1e-6,
		float	inferenceRate = 0.1f,
		float	precisionRate = 0.01f,
		float	lmbda = 1e-2f,
		ActivationType	aType = ActivationType::RELU,
		ActivationType	dType = ActivationType::dRELU )

Constructor for a Deepity Layer, using a named ActivationType instead of raw function pointers.

Parameters

size	Size of layer
nextSize	Size of next layer (used for row size of W)
batchSize	Batch size (default to 1 for simplicity)
learningRate	Learning rate to update weights
inferenceRate	Learning rate to update state
precisionRate	Precition weight to update precision
lmbda	Weight decay (L2 regularization) coefficient
aType	Activation type
dType	Activation derivative type

Member Function Documentation

◆ BindMemory()

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

Binds all tensor pointers to the contiguous memory block.

◆ CalculateState()

float Deep::DiscriminativePCLayer::CalculateState ( )

overridevirtualnoexcept

Calculates the total network energy state.

\[E = \sum_l 1/2 ||z^{(l)} - \mu^{(l)}||^2 \]

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

Implements Deep::Layer.

◆ ClampState()

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

noexcept

Clamps the layer to the input data.

Parameters

inputData Input

◆ ComputeMuOnly()

void Deep::DiscriminativePCLayer::ComputeMuOnly ( )

noexcept

Computes only mu (forward prediction), skipping error and energy computation entirely. Useful for callers that only need a forward pass through current weights (e.g. forward projection initialization) without the cost of the discarded error/energy values.

◆ Flush()

void Deep::DiscriminativePCLayer::Flush ( )

inlineoverridevirtualnoexcept

Does nothing; exists for class extension.

Reimplemented from Deep::Layer.

◆ GetActivationType()

ActivationType Deep::DiscriminativePCLayer::GetActivationType ( ) const

inlinenoexcept

Returns this layer's configured activation type.

Returns: ActivationType

◆ GetBatchSize()

size_t Deep::DiscriminativePCLayer::GetBatchSize ( ) const

inlineoverridevirtualnoexcept

Returns batchSize.

Returns: size_t batchSize

Implements Deep::Layer.

◆ GetBeliefs()

float * Deep::DiscriminativePCLayer::GetBeliefs ( )

inlineoverridevirtualnoexcept

Returns beliefs.

Returns: float *z

Implements Deep::Layer.

◆ GetBiases() [1/2]

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

inlinenoexcept

Returns a read-only version of the stored biases.

Returns: const float *b

◆ GetBiases() [2/2]

float * Deep::DiscriminativePCLayer::GetBiases ( )

inlinenoexcept

Returns a mutable version of the stored biases.

Returns: float *b

◆ GetDerivativeType()

ActivationType Deep::DiscriminativePCLayer::GetDerivativeType ( ) const

inlinenoexcept

Returns this layer's configured activation-derivative type.

Returns: ActivationType

◆ GetErrors()

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

inlineoverridevirtualnoexcept

Returns errors.

Returns: float *e

Implements Deep::Layer.

◆ GetInferenceRate()

float Deep::DiscriminativePCLayer::GetInferenceRate ( ) const

inlinenoexcept

Returns the inference rate used for state updates.

Returns: float ir

◆ GetInputSize()

size_t Deep::DiscriminativePCLayer::GetInputSize ( ) const

inlineoverridevirtualnoexcept

Returns size (input size).

Returns: size_t size

Implements Deep::Layer.

◆ GetLambda()

float Deep::DiscriminativePCLayer::GetLambda ( ) const

inlinenoexcept

Returns the weight-decay (L2 regularization) coefficient.

Returns: float lmbda

◆ GetLayerAbove()

const DiscriminativePCLayer & Deep::DiscriminativePCLayer::GetLayerAbove ( ) const

inlinenoexcept

Gets const pointer to layer above.

Returns: layerAbove

◆ GetLayerBelow()

const DiscriminativePCLayer & Deep::DiscriminativePCLayer::GetLayerBelow ( ) const

inlinenoexcept

Gets const pointer to layer below.

Returns: layerBelow

◆ GetLearningRate()

float Deep::DiscriminativePCLayer::GetLearningRate ( ) const

inlinenoexcept

Returns the learning rate used for weight updates.

Returns: float lr

◆ GetMu()

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

inlinenoexcept

Returns this layer's outgoing prediction buffer.

Returns: const float *mu

◆ GetOutputSize()

size_t Deep::DiscriminativePCLayer::GetOutputSize ( ) const

inlineoverridevirtualnoexcept

Returns nextSize (output size).

Returns: size_t nextSize

Implements Deep::Layer.

◆ GetPrecisionRate()

float Deep::DiscriminativePCLayer::GetPrecisionRate ( ) const

inlinenoexcept

Returns the learning rate used for precision updates.

Returns: float pr

◆ GetPrecisions()

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

inline

Returns a read-only version of the stored precisions.

Returns: const float *p

◆ GetRequiredFloats()

size_t Deep::DiscriminativePCLayer::GetRequiredFloats ( ) const

noexcept

Calculates exact float count required by this layer's architecture.

◆ GetWeights() [1/2]

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

inlinenoexcept

Returns a read-only version of the stored weights.

Returns: const float *W

◆ GetWeights() [2/2]

float * Deep::DiscriminativePCLayer::GetWeights ( )

inlinenoexcept

Returns a mutable version of the stored weights.

Returns: float *W

◆ RandomizeWeights()

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

noexcept

Makes the weights W randomized to [0.0, 0.1] using an OpenMP-parallelized uniform real distribution.

Parameters

twister The classic Mersenne Twister

◆ ResetState()

void Deep::DiscriminativePCLayer::ResetState ( )

noexcept

Resets 'z' (beliefs) to 0.

◆ ResyncLogPrecision()

void Deep::DiscriminativePCLayer::ResyncLogPrecision ( )

noexcept

Recomputes log of precision if it falls behind.

◆ SetInferenceRate()

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

inlinenoexcept

Sets the inference rate used for state updates.

Parameters

ir	The new inference rate.

◆ SetLambda()

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

inlinenoexcept

Sets the weight-decay (L2 regularization) coefficient.

Parameters

l	The new lambda value.

◆ SetLayerAbove()

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

inlinenoexcept

Ties this layer to one above it.

Parameters

above DiscriminativePCLayer*

◆ SetLayerBelow()

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

inlinenoexcept

Ties this layer to one below it.

Parameters

below DiscriminativePCLayer*

◆ SetLearningRate()

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

inlinenoexcept

Sets the learning rate used for weight updates.

Parameters

lr	The new learning rate.

◆ SetOptimizer()

void Deep::DiscriminativePCLayer::SetOptimizer ( OptimizerType o )

inlinenoexcept

Selects the optimizer used for weight updates.

Parameters

o	The optimizer type to use.

◆ SetPrecisionRate()

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

inlinenoexcept

Sets the learning rate used for precision updates.

Parameters

pr	The new precision rate.

◆ UnclampState()

void Deep::DiscriminativePCLayer::UnclampState ( )

noexcept

Unclamps the layer to the input data.

◆ UpdatePrecision()

void Deep::DiscriminativePCLayer::UpdatePrecision ( )

noexcept

Updates this layer's precision weighting based on the current prediction error.

◆ UpdateState()

void Deep::DiscriminativePCLayer::UpdateState ( )

overridevirtualnoexcept

Computes the state derivatives for inference.

\[\frac{dz^{(l)}}{dt} = -e^{(l)} + (W^{(l-1)})^T e^{(l-1)} \odot \sigma'(W^{(l-1)}z^{(l)}) \]

Implements Deep::Layer.

◆ UpdateWeights()

void Deep::DiscriminativePCLayer::UpdateWeights ( )

overridevirtualnoexcept

Computes weight updates via gradient descent, with L2 weight decay.

\[W^{(l)} \leftarrow (1 - \lambda) W^{(l)} - \eta e^{(l)} (z^{(l+1)})^T \]

Implements Deep::Layer.

◆ PCNDiagnostics

friend class PCNDiagnostics

friend

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

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

Public Member Functions

Friends

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ DiscriminativePCLayer() [1/2]

◆ DiscriminativePCLayer() [2/2]

Member Function Documentation

◆ BindMemory()

◆ CalculateState()

◆ ClampState()

◆ ComputeMuOnly()

◆ Flush()

◆ GetActivationType()

◆ GetBatchSize()

◆ GetBeliefs()

◆ GetBiases() [1/2]

◆ GetBiases() [2/2]

◆ GetDerivativeType()

◆ GetErrors()

◆ GetInferenceRate()

◆ GetInputSize()

◆ GetLambda()

◆ GetLayerAbove()

◆ GetLayerBelow()

◆ GetLearningRate()

◆ GetMu()

◆ GetOutputSize()

◆ GetPrecisionRate()

◆ GetPrecisions()

◆ GetRequiredFloats()

◆ GetWeights() [1/2]

◆ GetWeights() [2/2]

◆ RandomizeWeights()

◆ ResetState()

◆ ResyncLogPrecision()

◆ SetInferenceRate()

◆ SetLambda()

◆ SetLayerAbove()

◆ SetLayerBelow()

◆ SetLearningRate()

◆ SetOptimizer()

◆ SetPrecisionRate()

◆ UnclampState()

◆ UpdatePrecision()

◆ UpdateState()

◆ UpdateWeights()

Friends And Related Symbol Documentation

◆ PCNDiagnostics