Class BaseLayer

Base class for the layers

Inheritance

System.Object

BaseLayer

Elu

Exp

HardSigmoid

LeakyRelu

Linear

PRelu

Relu

Sigmoid

Softmax

Softplus

Softsign

Tanh

AvgPooling1D

AvgPooling2D

AvgPooling3D

BatchNormalization

Conv1D

Conv2D

Conv2DTranspose

Conv3D

Conv3DTranspose

Dense

Dropout

Embedding

Flatten

GlobalPooling1D

GlobalPooling2D

GlobalPooling3D

MaxPooling1D

MaxPooling2D

MaxPooling3D

Permute

Repeat

Reshape

Namespace: SiaNet.Layers

Assembly: SiaNet.dll

Syntax

public abstract class BaseLayer

Constructors

BaseLayer(String)

Initializes a new instance of the BaseLayer class.

Declaration

public BaseLayer(string name)

Parameters

Type	Name	Description
System.String	name	The name.

Fields

Input

The input tensor parameter

Declaration

[NonSerialized]
public Parameter Input

Field Value

Type	Description
Parameter

Output

The output tensor parameter

Declaration

[NonSerialized]
public Tensor Output

Field Value

Type	Description
Tensor

Params

The parameters list used by the layer

Declaration

[NonSerialized]
public Dictionary<string, Parameter> Params

Field Value

Type	Description
System.Collections.Generic.Dictionary<System.String, Parameter>

Properties

Item[String]

Gets or sets the Parameter with the specified name.

Declaration

public Parameter this[string name] { get; set; }

Parameters

Type	Name	Description
System.String	name	The name.

Property Value

Type	Description
Parameter	The Parameter.

Name

Gets or sets the name of the layer

Declaration

public string Name { get; set; }

Property Value

Type	Description
System.String	The name.

SkipPred

Gets or sets a value indicating whether the layer is train only

Declaration

public bool SkipPred { get; set; }

Property Value

Type	Description
System.Boolean	true if [skip pred]; otherwise, false.

Methods

Backward(Tensor)

Calculate the gradient of this layer function

Declaration

public virtual void Backward(Tensor outputgrad)

Parameters

Type	Name	Description
Tensor	outputgrad	The calculated output grad from previous layer.

BuildParam(String, Int64[], DataType, BaseInitializer, BaseConstraint, BaseRegularizer, Boolean)

Builds the parameter with specified parameter. Used to create weight or bias parameter for the layer.

Declaration

public Parameter BuildParam(string name, long[] shape, DataType elementType, BaseInitializer initializer, BaseConstraint constraint = null, BaseRegularizer regularizer = null, bool trainable = true)

Parameters

Type	Name	Description
System.String	name	The name of the parameter.
System.Int64[]	shape	The shape of the tensor.
DataType	elementType	Datatype of the tensor.
BaseInitializer	initializer	The initializer used to create the parameter.
BaseConstraint	constraint	The constraint function applied for this parameter.
BaseRegularizer	regularizer	The regularizer function for this parameter.
System.Boolean	trainable	if set to true [trainable].

Returns

Type	Description
Parameter

Forward(Tensor)

Forwards the inputs and compute the output

Declaration

public virtual void Forward(Tensor x)

Parameters

Type	Name	Description
Tensor	x	The input tensor for this layer.