boone.structure

Class ConvolutionLayer

java.lang.Object

boone.structure.Layer

boone.structure.ConvolutionLayer

All Implemented Interfaces:

Storable

public class ConvolutionLayer
extends Layer

A convolution layer. The depth of the layer is the number of filters, i.e., outgoing channels. $Id$

Since:

October, 2021

Author:

Helmut A. Mayer

Constructor Summary

Constructors

Constructor and Description
ConvolutionLayer() Constructs an empty convolution layer with all zero values and default neurons and links.
ConvolutionLayer(int filterSize, int padSize, int strideSize, int channels) Constructs a convolution layer with parameters having identical width and height, a default ReLU neuron, and a default link.
ConvolutionLayer(int filterWidth, int filterHeight, int padWidth, int padHeight, int strideWidth, int strideHeight, int channels, Neuron neuron, Link link) Constructs a general convolution layer.
ConvolutionLayer(int filterSize, int padSize, int strideSize, int channels, Neuron neuron, Link link) Constructs a convolution layer with parameters having identical width and height, and given neuron and link type.

Method Summary

Modifier and Type	Method and Description
void	addMap(ConvolutionMap map) Adds a convolution map to this layer.
java.lang.String	toString() Returns the width x height x depth of this layer.

Methods inherited from class boone.structure.Layer

addMap, addMap, connect, disconnect, fromXML, getDepth, getHeight, getLevel, getLinkCount, getMapCount, getMaps, getNeuron, getNeurons, getSize, getWidth, removeMap, setIOType, setLevel, toXML

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Detail

ConvolutionLayer

public ConvolutionLayer()

Constructs an empty convolution layer with all zero values and default neurons and links.

ConvolutionLayer

public ConvolutionLayer(int filterSize,
                        int padSize,
                        int strideSize,
                        int channels)

Constructs a convolution layer with parameters having identical width and height, a default ReLU neuron, and a default link.

Parameters:

filterSize - the width and height of the filter

padSize - the width and height of the pad

strideSize - the width and height of the stride

channels - the number of filters

ConvolutionLayer

public ConvolutionLayer(int filterSize,
                        int padSize,
                        int strideSize,
                        int channels,
                        Neuron neuron,
                        Link link)

Constructs a convolution layer with parameters having identical width and height, and given neuron and link type.

Parameters:

filterSize - the width and height of the filter

padSize - the width and height of the pad

strideSize - the width and height of the stride

channels - the number of filters

neuron - the neuron type (if null, default neuron)

link - the link type (if null, default link)

ConvolutionLayer

public ConvolutionLayer(int filterWidth,
                        int filterHeight,
                        int padWidth,
                        int padHeight,
                        int strideWidth,
                        int strideHeight,
                        int channels,
                        Neuron neuron,
                        Link link)

Constructs a general convolution layer. Note that after construction the layer contains no neurons, as they are determined when connecting the layer.

Parameters:

filterWidth - the width of the filter

filterHeight - the height of the filter

padWidth - the width of the pad

padHeight - the height of the pad

strideWidth - the width of the stride

strideHeight - the height of the stride

channels - the number of filters

neuron - the neuron type (if null, default neuron)

link - the link type (if null, default link)

Method Detail

addMap

public void addMap(ConvolutionMap map)

Adds a convolution map to this layer.

Parameters:

map - a convolution map

toString

public java.lang.String toString()

Returns the width x height x depth of this layer.

Overrides:

toString in class Layer

Returns:

the string expression