org.cognity.neural.net

Class MlpNet.Builder

java.lang.Object

org.cognity.neural.net.MlpNet.Builder

All Implemented Interfaces:

NeuralNetBuilder<MlpNet>

Enclosing class:

MlpNet

public static final class MlpNet.Builder
extends Object
implements NeuralNetBuilder<MlpNet>

A builder for a multilayer perceptron. In order to build a network, perform the following steps:

1. set the input size by calling setInputSize(int)
2. add layers to the network using the add* methods
3. obtain the network object by calling build method

The last added layer automatically becomes the output layer. Layers are connected by addition order.
The network is initialized using UniformInitializer class.

Author:

Artur Markiewicz

Constructor Summary

Constructors

Constructor and Description
MlpNet.Builder() Creates a new instance of the builder.

Method Summary

Methods

Modifier and Type	Method and Description
MlpNet.Builder	addCustom(int size, Function activationFun) Adds a layer with specified size and activation function.
MlpNet.Builder	addLinear(int size) Adds a layer with specified size and identity function as the activation function.
MlpNet.Builder	addSigmoid(int size) Adds a layer with specified size and sigmoid function as the activation function.
MlpNet.Builder	addStep(int size) Adds a layer with specified size and Heaviside step function as the activation function.
MlpNet.Builder	addTanh(int size) Adds a layer with specified size and hyperbolic tangent as the activation function.
MlpNet	build() Returns a new multilayer perceptron.
MlpNet.Builder	setInputSize(int size) Sets the number of inputs for a network.

Methods inherited from class java.lang.Object

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

Constructor Detail

MlpNet.Builder

public MlpNet.Builder()

Creates a new instance of the builder.

Method Detail

setInputSize

public MlpNet.Builder setInputSize(int size)

Sets the number of inputs for a network.

Parameters:

size - the number of inputs. Must be greater than zero.

Returns:

this builder

addCustom

public MlpNet.Builder addCustom(int size,
                       Function activationFun)

Adds a layer with specified size and activation function.

Parameters:

size - the layer's size. Must be greater than zero.

activationFun - the activation function. Cannot be null.

Returns:

this builder

addTanh

public MlpNet.Builder addTanh(int size)

Adds a layer with specified size and hyperbolic tangent as the activation function.

Parameters:

size - the layer's size. Must be greater than zero.

Returns:

this builder

addSigmoid

public MlpNet.Builder addSigmoid(int size)

Adds a layer with specified size and sigmoid function as the activation function.

Parameters:

size - the layer's size. Must be greater than zero.

Returns:

this builder

addLinear

public MlpNet.Builder addLinear(int size)

Adds a layer with specified size and identity function as the activation function.

Parameters:

size - the layer's size. Must be greater than zero.

Returns:

this builder

addStep

public MlpNet.Builder addStep(int size)

Adds a layer with specified size and Heaviside step function as the activation function.

Parameters:

size - the layer's size. Must be greater than zero.

Returns:

this builder

build

public MlpNet build()

Returns a new multilayer perceptron. The perceptron is already initialized using an uniform initializer.

Specified by:

build in interface NeuralNetBuilder<MlpNet>

Returns:

a new multilayer perceptron