PC layer with Gauss-Seidel (sequential-sweep) settling dynamics, in place of the usual simultaneous update. Not yet gradient-checked – see the file-level warning above. More...

#include <GaussSeidelPCLayer.h>

Inheritance diagram for Deep::GaussSeidelPCLayer:

Public Member Functions
	GaussSeidelPCLayer (int size, int nextSize, int batchSize=1, float learningRate=1e-6f, float inferenceRate=0.1f, float lmbda=1e-2f, void(act)(float , size_t)=relu, void(dAct)(float , size_t, bool)=dRelu)
	Constructor for a GaussSeidelPCLayer.
	GaussSeidelPCLayer (int size, int nextSize, int batchSize=1, float learningRate=1e-6f, float inferenceRate=0.1f, float lmbda=1e-2f, ActivationType aType=ActivationType::RELU, ActivationType dType=ActivationType::dRELU)
	Constructor using a named ActivationType instead of raw function pointers.
float	CalculateState () noexcept override
	Step 1 of a Gauss-Seidel timestep: updates z using mu/e_above HELD OVER from the end of the previous timestep. Does NOT mutate mu – computes its derivative into a separate scratch buffer.
void	UpdateState () noexcept override
	Step 1: the actual z-update. See class-level docs for the required three-sweep calling order.
void	ComputePrediction () noexcept
	Step 2 of a Gauss-Seidel timestep: recomputes mu fresh, using the z JUST updated by UpdateState() in this same timestep. Call on EVERY layer before any layer's ComputeError().
float	ComputeError () noexcept
	Step 3 of a Gauss-Seidel timestep: recomputes e fresh, using this layer's own (just-updated) z as the target and layerBelow's FRESH mu (from its ComputePrediction() call) as the prediction.
void	UpdateWeights () noexcept override
	Computes weight updates via gradient descent, with L2 weight decay. Called once after the full settling loop completes, matching ngc-learn's evolve_process.
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.
void	UnclampState () noexcept
	Releases a previous ClampState() call.
float *	GetBeliefs () noexcept override
	Returns this layer's belief buffer.
const float *	GetMu () const noexcept
	Returns this layer's outgoing prediction buffer.
const float *	GetErrors () const noexcept override
	Returns this layer's prediction-error buffer.
size_t	GetInputSize () const noexcept override
	Returns this layer's own belief size.
size_t	GetOutputSize () const noexcept override
	Returns this layer's outgoing prediction size (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 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.
float	GetLearningRate () const noexcept
	Returns the learning rate used for weight updates.
float	GetInferenceRate () const noexcept
	Returns the inference rate (Euler integration step size).
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 (Euler integration step size).
void	SetLambda (float l) noexcept
	Sets the weight-decay (L2 regularization) coefficient.
void	SetOptimizer (const OptimizerType o) noexcept
	Selects the optimizer used for weight updates.
void	SetLayerAbove (GaussSeidelPCLayer *above) noexcept
	Sets the layer immediately above this one in the network.
void	SetLayerBelow (GaussSeidelPCLayer *below) noexcept
	Sets the layer immediately below this one in the network.
void	ResetState () noexcept
	Resets beliefs/errors/predictions back to their initial values, without touching learned weights.
const GaussSeidelPCLayer &	GetLayerAbove () const noexcept
	Returns the layer immediately above this one.
const GaussSeidelPCLayer &	GetLayerBelow () const noexcept
	Returns the layer immediately below this one.
void	RandomizeWeights (std::mt19937 &twister) noexcept
	Randomizes this layer's weights (and the E feedback-alignment matrix) in place.
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
	Computes the total number of floats this layer requires from a MemoryArena.
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
virtual std::map< std::string, TensorDescriptor >	GetStateDict () const
	Returns a dictionary of this.

Friends
class	GaussSeidelPCNDiagnostics

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

Detailed Description

PC layer with Gauss-Seidel (sequential-sweep) settling dynamics, in place of the usual simultaneous update. Not yet gradient-checked – see the file-level warning above.

Constructor & Destructor Documentation

◆ GaussSeidelPCLayer() [1/2]

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

Constructor for a GaussSeidelPCLayer.

Parameters

size	Size of this layer's own belief (z)
nextSize	Size of the layer above's belief (this layer's outgoing prediction target); 0 marks a terminal layer
batchSize	Batch size
learningRate	Learning rate for weight updates
inferenceRate	Inference rate (Euler integration step size)
lmbda	Weight decay (L2 regularization) coefficient
act	Activation function
dAct	Derivative of activation function

◆ GaussSeidelPCLayer() [2/2]

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

Constructor using a named ActivationType instead of raw function pointers.

Parameters

size	Size of this layer's own belief (z)
nextSize	Size of the layer above's belief (this layer's outgoing prediction target); 0 marks a terminal layer
batchSize	Batch size
learningRate	Learning rate for weight updates
inferenceRate	Inference rate (Euler integration step size)
lmbda	Weight decay (L2 regularization) coefficient
aType	Activation type
dType	Activation derivative type

Member Function Documentation

◆ BindMemory()

void Deep::GaussSeidelPCLayer::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::GaussSeidelPCLayer::CalculateState ( )

inlineoverridevirtualnoexcept

Step 1 of a Gauss-Seidel timestep: updates z using mu/e_above HELD OVER from the end of the previous timestep. Does NOT mutate mu – computes its derivative into a separate scratch buffer.

Returns: Always 0.0f – energy is only meaningful from ComputeError(), kept for Layer interface conformance.

Implements Deep::Layer.

◆ ClampState()

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

noexcept

Clamps this layer's beliefs to externally-provided data.

Parameters

inputData Flattened input data of length size.

◆ ComputeError()

float Deep::GaussSeidelPCLayer::ComputeError ( )

noexcept

Step 3 of a Gauss-Seidel timestep: recomputes e fresh, using this layer's own (just-updated) z as the target and layerBelow's FRESH mu (from its ComputePrediction() call) as the prediction.

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

◆ ComputePrediction()

void Deep::GaussSeidelPCLayer::ComputePrediction ( )

noexcept

Step 2 of a Gauss-Seidel timestep: recomputes mu fresh, using the z JUST updated by UpdateState() in this same timestep. Call on EVERY layer before any layer's ComputeError().

◆ Flush()

void Deep::GaussSeidelPCLayer::Flush ( )

inlineoverridevirtualnoexcept

No-op; exists for Layer interface conformance.

Reimplemented from Deep::Layer.

◆ GetActivationType()

ActivationType Deep::GaussSeidelPCLayer::GetActivationType ( ) const

inlinenoexcept

Returns this layer's configured activation type.

Returns: ActivationType

◆ GetBatchSize()

size_t Deep::GaussSeidelPCLayer::GetBatchSize ( ) const

inlineoverridevirtualnoexcept

Returns the batch size this layer was constructed with.

Returns: size_t batchSize

Implements Deep::Layer.

◆ GetBeliefs()

float * Deep::GaussSeidelPCLayer::GetBeliefs ( )

inlineoverridevirtualnoexcept

Returns this layer's belief buffer.

Returns: Pointer to this layer's beliefs.

Implements Deep::Layer.

◆ GetBiases() [1/2]

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

inlinenoexcept

Returns a read-only version of the stored biases.

Returns: const float *b

◆ GetBiases() [2/2]

float * Deep::GaussSeidelPCLayer::GetBiases ( )

inlinenoexcept

Returns a mutable version of the stored biases.

Returns: float *b

◆ GetDerivativeType()

ActivationType Deep::GaussSeidelPCLayer::GetDerivativeType ( ) const

inlinenoexcept

Returns this layer's configured activation-derivative type.

Returns: ActivationType

◆ GetErrors()

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

inlineoverridevirtualnoexcept

Returns this layer's prediction-error buffer.

Returns: const float *e

Implements Deep::Layer.

◆ GetInferenceRate()

float Deep::GaussSeidelPCLayer::GetInferenceRate ( ) const

inlinenoexcept

Returns the inference rate (Euler integration step size).

Returns: float ir

◆ GetInputSize()

size_t Deep::GaussSeidelPCLayer::GetInputSize ( ) const

inlineoverridevirtualnoexcept

Returns this layer's own belief size.

Returns: size_t size

Implements Deep::Layer.

◆ GetLambda()

float Deep::GaussSeidelPCLayer::GetLambda ( ) const

inlinenoexcept

Returns the weight-decay (L2 regularization) coefficient.

Returns: float lmbda

◆ GetLayerAbove()

const GaussSeidelPCLayer & Deep::GaussSeidelPCLayer::GetLayerAbove ( ) const

inlinenoexcept

Returns the layer immediately above this one.

Warning: Dereferences layerAbove without a null check; only valid if SetLayerAbove() was previously called with a non-null pointer.

◆ GetLayerBelow()

const GaussSeidelPCLayer & Deep::GaussSeidelPCLayer::GetLayerBelow ( ) const

inlinenoexcept

Returns the layer immediately below this one.

Warning: Dereferences layerBelow without a null check; only valid if SetLayerBelow() was previously called with a non-null pointer.

◆ GetLearningRate()

float Deep::GaussSeidelPCLayer::GetLearningRate ( ) const

inlinenoexcept

Returns the learning rate used for weight updates.

Returns: float lr

◆ GetMu()

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

inlinenoexcept

Returns this layer's outgoing prediction buffer.

Returns: const float *mu

◆ GetOutputSize()

size_t Deep::GaussSeidelPCLayer::GetOutputSize ( ) const

inlineoverridevirtualnoexcept

Returns this layer's outgoing prediction size (0 for a terminal layer).

Returns: size_t nextSize

Implements Deep::Layer.

◆ GetRequiredFloats()

size_t Deep::GaussSeidelPCLayer::GetRequiredFloats ( ) const

noexcept

Computes the total number of floats this layer requires from a MemoryArena.

Returns: The required float count.

◆ GetWeights() [1/2]

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

inlinenoexcept

Returns a read-only version of the stored weights.

Returns: const float *W

◆ GetWeights() [2/2]

float * Deep::GaussSeidelPCLayer::GetWeights ( )

inlinenoexcept

Returns a mutable version of the stored weights.

Returns: float *W

◆ RandomizeWeights()

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

noexcept

Randomizes this layer's weights (and the E feedback-alignment matrix) in place.

Parameters

twister The classic Mersenne Twister

◆ ResetState()

void Deep::GaussSeidelPCLayer::ResetState ( )

noexcept

Resets beliefs/errors/predictions back to their initial values, without touching learned weights.

◆ SetInferenceRate()

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

inlinenoexcept

Sets the inference rate (Euler integration step size).

Parameters

ir	The new inference rate.

◆ SetLambda()

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

inlinenoexcept

Sets the weight-decay (L2 regularization) coefficient.

Parameters

l	The new lambda value.

◆ SetLayerAbove()

void Deep::GaussSeidelPCLayer::SetLayerAbove ( GaussSeidelPCLayer * 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::GaussSeidelPCLayer::SetLayerBelow ( GaussSeidelPCLayer * 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::GaussSeidelPCLayer::SetLearningRate ( float lr )

inlinenoexcept

Sets the learning rate used for weight updates.

Parameters

lr	The new learning rate.

◆ SetOptimizer()

void Deep::GaussSeidelPCLayer::SetOptimizer ( const OptimizerType o )

inlinenoexcept

Selects the optimizer used for weight updates.

Parameters

o	The optimizer type to use.

◆ UnclampState()

void Deep::GaussSeidelPCLayer::UnclampState ( )

noexcept

Releases a previous ClampState() call.

◆ UpdateState()

void Deep::GaussSeidelPCLayer::UpdateState ( )

overridevirtualnoexcept

Step 1: the actual z-update. See class-level docs for the required three-sweep calling order.

Implements Deep::Layer.

◆ UpdateWeights()

void Deep::GaussSeidelPCLayer::UpdateWeights ( )

overridevirtualnoexcept

Computes weight updates via gradient descent, with L2 weight decay. Called once after the full settling loop completes, matching ngc-learn's evolve_process.

Implements Deep::Layer.

◆ GaussSeidelPCNDiagnostics

friend class GaussSeidelPCNDiagnostics

friend

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

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

Public Member Functions

Friends

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ GaussSeidelPCLayer() [1/2]

◆ GaussSeidelPCLayer() [2/2]

Member Function Documentation

◆ BindMemory()

◆ CalculateState()

◆ ClampState()

◆ ComputeError()

◆ ComputePrediction()

◆ Flush()

◆ GetActivationType()

◆ GetBatchSize()

◆ GetBeliefs()

◆ GetBiases() [1/2]

◆ GetBiases() [2/2]

◆ GetDerivativeType()

◆ GetErrors()

◆ GetInferenceRate()

◆ GetInputSize()

◆ GetLambda()

◆ GetLayerAbove()

◆ GetLayerBelow()

◆ GetLearningRate()

◆ GetMu()

◆ GetOutputSize()

◆ GetRequiredFloats()

◆ GetWeights() [1/2]

◆ GetWeights() [2/2]

◆ RandomizeWeights()

◆ ResetState()

◆ SetInferenceRate()

◆ SetLambda()

◆ SetLayerAbove()

◆ SetLayerBelow()

◆ SetLearningRate()

◆ SetOptimizer()

◆ UnclampState()

◆ UpdateState()

◆ UpdateWeights()

Friends And Related Symbol Documentation

◆ GaussSeidelPCNDiagnostics