goneural

package module
v0.0.4 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Nov 25, 2021 License: GPL-3.0 Imports: 5 Imported by: 0

README

goneural

Golang neural networking

Development package

This is a development package and currently only supports dense layers. There is also currently no way to train the networks.

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func GetLayerTotLoss 

func GetLayerTotLoss(l LossFunction, expectedX, predictedX *Matrix) float64

func ShuffleBoth 

func ShuffleBoth(X, Y []*Matrix)

Types

type Activation 

type Activation interface {
	// Calc : Calculate the activation for x
	Calc(x float64) float64
	// Diff : Calculate the differential of the activation for x
	Diff(x float64) float64
}

Activation : An interface type that denotes an activations function.

type DenseLayer 

type DenseLayer struct {
	Weights Matrix
	Ac      Activation
	// contains filtered or unexported fields
}

DenseLayer : The standard type of dense neural network layer

func NewDenseLayer 

func NewDenseLayer(numInputs, numOutputs int, activation Activation) *DenseLayer

func (*DenseLayer) GetActivation 

func (l *DenseLayer) GetActivation() Activation

func (*DenseLayer) GetNumInputs 

func (l *DenseLayer) GetNumInputs() int

func (*DenseLayer) GetNumOutputs 

func (l *DenseLayer) GetNumOutputs() int

func (*DenseLayer) PropagateValues 

func (l *DenseLayer) PropagateValues(X *Matrix) *Matrix

func (*DenseLayer) TrainGradientDescent 

func (l *DenseLayer) TrainGradientDescent(learningRate float64, layerInputs, deltas *Matrix) *Matrix

TrainGradientDescent : DEPRECATED

func (*DenseLayer) TrainMinibatch 

func (l *DenseLayer) TrainMinibatch(learningRate float64, avInputs, avDeltas *Matrix) *Matrix

type FeedForwardNetwork 

type FeedForwardNetwork struct {
	Inputs  int
	Outputs int
	Layers  []Layer
	Loss    LossFunction
}

func NewFeedForwardNetwork 

func NewFeedForwardNetwork(inputs int, loss LossFunction) *FeedForwardNetwork

func (*FeedForwardNetwork) AddLayer 

func (n *FeedForwardNetwork) AddLayer(layer Layer)

func (*FeedForwardNetwork) GetLastLayerDeltas 

func (n *FeedForwardNetwork) GetLastLayerDeltas(pred, expec *Matrix) *Matrix

func (*FeedForwardNetwork) Predict 

func (n *FeedForwardNetwork) Predict(X *Matrix) *Matrix

func (*FeedForwardNetwork) PredictAll 

func (n *FeedForwardNetwork) PredictAll(Xs []*Matrix) []*Matrix

type Layer 

type Layer interface {
	// PropagateValues : Propagates an input matrix though the layer and returns the calculated matrix
	PropagateValues(X *Matrix) *Matrix
	// GetNumInputs : Gets the number of inputs
	GetNumInputs() int
	// GetNumOutputs : Gets the number of outputs
	GetNumOutputs() int
	// GetActivation : Gets the activation function of this layer
	GetActivation() Activation
}

Layer : A interface type that describes layer in a feed forward network

type LeakyReLUActivation 

type LeakyReLUActivation struct {
	// SubZeroGradient : The gradient of the function when x < 0
	SubZeroGradient float64
}

LeakyReLUActivation : A leaky ReLU activation

func (LeakyReLUActivation) Calc 

func (a LeakyReLUActivation) Calc(x float64) float64

func (LeakyReLUActivation) Diff 

func (a LeakyReLUActivation) Diff(x float64) float64

type LinearActivation 

type LinearActivation types.Nil

LinearActivation : Activation where y=x

func (LinearActivation) Calc 

func (a LinearActivation) Calc(x float64) float64

func (LinearActivation) Diff 

func (a LinearActivation) Diff(x float64) float64

type LossFunction 

type LossFunction interface {
	SingleLoss(expected, predicted float64) float64
	SingleLossDiff(expected, predicted float64) float64
}

type Matrix 

type Matrix struct {
	Shape  []int
	Values []float64
}

func GetLayerLossDiffs 

func GetLayerLossDiffs(l LossFunction, expectedX, predictedX *Matrix) *Matrix

func New1DMatrix 

func New1DMatrix(vals []float64) *Matrix

func New2DMatrix 

func New2DMatrix(vals [][]float64) *Matrix

func NewZerosMatrix 

func NewZerosMatrix(shape ...int) *Matrix

func (*Matrix) Add 

func (m *Matrix) Add(m2 *Matrix) *Matrix

func (*Matrix) AddValue1D 

func (m *Matrix) AddValue1D(i int, v float64)

func (*Matrix) AddValue2D 

func (m *Matrix) AddValue2D(x, y int, v float64)

func (*Matrix) Copy 

func (m *Matrix) Copy() *Matrix

func (*Matrix) CopyTo 

func (m *Matrix) CopyTo(m2 *Matrix)

func (*Matrix) CopyToSmaller 

func (m *Matrix) CopyToSmaller(m2 *Matrix)

func (*Matrix) Get1DLength 

func (m *Matrix) Get1DLength() int

func (*Matrix) GetValue1D 

func (m *Matrix) GetValue1D(i int) float64

func (*Matrix) GetValue2D 

func (m *Matrix) GetValue2D(x, y int) float64

func (*Matrix) Mul 

func (m *Matrix) Mul(f float64) *Matrix

func (*Matrix) SetValue1D 

func (m *Matrix) SetValue1D(i int, v float64)

func (*Matrix) SetValue2D 

func (m *Matrix) SetValue2D(x, y int, v float64)

type MinibatchOptimizer 

type MinibatchOptimizer struct {
	LearningRate float64
	BatchSize    int
}

func (*MinibatchOptimizer) FitBatch 

func (o *MinibatchOptimizer) FitBatch(n *FeedForwardNetwork, X, Y []*Matrix)

func (*MinibatchOptimizer) GetBatchSize 

func (o *MinibatchOptimizer) GetBatchSize() int

type Optimizer 

type Optimizer interface {
	FitBatch(n *FeedForwardNetwork, X, Y []*Matrix)
	GetBatchSize() int
}

type ReLUActivation 

type ReLUActivation types.Nil

ReLUActivation : An activation for ReLU

func (ReLUActivation) Calc 

func (a ReLUActivation) Calc(x float64) float64

func (ReLUActivation) Diff 

func (a ReLUActivation) Diff(x float64) float64

type SGDLayer 

type SGDLayer interface {
	Layer
	TrainMinibatch(learningRate float64, layerInputs, totDeltas *Matrix) *Matrix
}

type SerialTrainer 

type SerialTrainer struct {
	Opt                     Optimizer
	UseRelativeLearningRate bool
}

func (*SerialTrainer) Fit 

func (s *SerialTrainer) Fit(n *FeedForwardNetwork, Xs, Ys []*Matrix, epochs, printRate int)

type SigmoidActivation 

type SigmoidActivation types.Nil

SigmoidActivation : A sigmoid activation

func (SigmoidActivation) Calc 

func (a SigmoidActivation) Calc(x float64) float64

func (SigmoidActivation) Diff 

func (a SigmoidActivation) Diff(x float64) float64

type SquaredErrorLoss 

type SquaredErrorLoss types.Nil

func (SquaredErrorLoss) SingleLoss 

func (l SquaredErrorLoss) SingleLoss(expected, predicted float64) float64

func (SquaredErrorLoss) SingleLossDiff 

func (l SquaredErrorLoss) SingleLossDiff(expected, predicted float64) float64

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.