convolutionalneuralnetwork.md

ReImproveJS

Globals / ConvolutionalNeuralNetwork /

Class: ConvolutionalNeuralNetwork

deprecated Do not use convolutional networks with ReImproveJS for now, they are not fully implemented and tested in the library.

Hierarchy

• NeuralNetwork

  • ConvolutionalNeuralNetwork

Index

Constructors

• constructor

Properties

• convolutionalLayers
• flattenLayer
• inputShape

Accessors

• FlattenLayer
• InputShape

Methods

• addConvolutionalLayer
• addConvolutionalLayers
• addMaxPooling2DLayer
• addNeuralNetworkLayer
• addNeuralNetworkLayers
• createLayers
• getLayers

Object literals

• DEFAULT_CONV_LAYER
• DEFAULT_POOLING_LAYER

Constructors

constructor

+ new ConvolutionalNeuralNetwork(): ConvolutionalNeuralNetwork

Overrides NeuralNetwork.constructor

Defined in reimprove/networks.ts:121

Returns: ConvolutionalNeuralNetwork

---

Properties

Private convolutionalLayers

● convolutionalLayers: ConvolutionalNetworkLayer[]

Defined in reimprove/networks.ts:107

---

Private flattenLayer

● flattenLayer: FlattenLayer

Defined in reimprove/networks.ts:108

---

Protected inputShape

● inputShape: number[]

Inherited from NeuralNetwork.inputShape

Defined in reimprove/networks.ts:55

---

Accessors

FlattenLayer

● set FlattenLayer(layer: FlattenLayer): void

Defined in reimprove/networks.ts:162

Parameters:

Name	Type
layer	FlattenLayer

Returns: void

---

InputShape

● set InputShape(shape: number[]): void

Inherited from NeuralNetwork.InputShape

Defined in reimprove/networks.ts:85

Parameters:

Name	Type
shape	number[]

Returns: void

---

Methods

addConvolutionalLayer

▸ addConvolutionalLayer(layer: number | ConvolutionalNetworkLayer): void

Defined in reimprove/networks.ts:133

Parameters:

Name	Type
layer	number | ConvolutionalNetworkLayer

Returns: void

---

addConvolutionalLayers

▸ addConvolutionalLayers(layers: Array<number | ConvolutionalNetworkLayer>): void

Defined in reimprove/networks.ts:146

Parameters:

Name	Type
layers	Array<number | ConvolutionalNetworkLayer>

Returns: void

---

addMaxPooling2DLayer

▸ addMaxPooling2DLayer(layer?: MaxPooling2DLayer): void

Defined in reimprove/networks.ts:129

Parameters:

Name	Type
layer?	MaxPooling2DLayer

Returns: void

---

addNeuralNetworkLayer

▸ addNeuralNetworkLayer(layer: number | NeuralNetworkLayer): void

Inherited from NeuralNetwork.addNeuralNetworkLayer

Defined in reimprove/networks.ts:69

Parameters:

Name	Type
layer	number | NeuralNetworkLayer

Returns: void

---

addNeuralNetworkLayers

▸ addNeuralNetworkLayers(layers: Array<number | NeuralNetworkLayer>): void

Inherited from NeuralNetwork.addNeuralNetworkLayers

Defined in reimprove/networks.ts:81

Parameters:

Name	Type
layers	Array<number | NeuralNetworkLayer>

Returns: void

---

createLayers

▸ createLayers(includeInputShape: boolean): Array<any>

Overrides NeuralNetwork.createLayers

Defined in reimprove/networks.ts:150

Parameters:

Name	Type	Default
includeInputShape	boolean	true

Returns: Array<any>

---

getLayers

▸ getLayers(): Layer[]

Overrides NeuralNetwork.getLayers

Defined in reimprove/networks.ts:166

Returns: Layer[]

---

Object literals

Static Private DEFAULT_CONV_LAYER

■ DEFAULT_CONV_LAYER: object

Defined in reimprove/networks.ts:110

activation

● activation: string = "relu"

Defined in reimprove/networks.ts:113

filters

● filters: number = 32

Defined in reimprove/networks.ts:111

kernelSize

● kernelSize: number = 3

Defined in reimprove/networks.ts:112

type

● type: "convolutional" = "convolutional"

Defined in reimprove/networks.ts:114

---

Static Private DEFAULT_POOLING_LAYER

■ DEFAULT_POOLING_LAYER: object

Defined in reimprove/networks.ts:117

poolSize

● poolSize: number = 2

Defined in reimprove/networks.ts:118

strides

● strides: null = null

Defined in reimprove/networks.ts:119

type

● type: "maxpooling" = "maxpooling"

Defined in reimprove/networks.ts:120

---