Package dk.alexandra.fresco.stat.mlp

Class MLP

Object

MLP

public class MLP
extends Object

This class represents layered neural networks. Computing a networks output on a sample is done using the apply(ArrayList) method. Prediction is done using Predict. Training a network is done using the fit(List, List, int, double) method. Instances of this class are immutable.

Constructor Summary

Constructors

Constructor	Description
MLP(List<Layer> layers)

Method Summary

Modifier and Type	Method	Description
Computation<ArrayList<DRes<SFixed>>,ProtocolBuilderNumeric>	apply(ArrayList<DRes<SFixed>> input)	Apply this neural network in an input vector.
Computation<MLP,ProtocolBuilderNumeric>	fit(List<ArrayList<DRes<SFixed>>> data, List<ArrayList<DRes<SFixed>>> labels, int epochs, double learningRate)	Train this neural network using single step training (batch size = 1) and return a new neural network with the updated weights.
Layer	getLayer(int i)

Methods inherited from class Object

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

Constructor Details

MLP

public MLP(List<Layer> layers)

Method Details

getLayer

public Layer getLayer(int i)

fit

public Computation<MLP,ProtocolBuilderNumeric> fit(List<ArrayList<DRes<SFixed>>> data, List<ArrayList<DRes<SFixed>>> labels, int epochs, double learningRate)

Train this neural network using single step training (batch size = 1) and return a new neural network with the updated weights.

Parameters:

data - The training data.

labels - The expected outputs.

epochs - The number of epochs to run the training. One epoch is an iteration over all training data.

learningRate - The learning rate.

Returns:

A new neural network with weights updated from this network.

apply

public Computation<ArrayList<DRes<SFixed>>,ProtocolBuilderNumeric> apply(ArrayList<DRes<SFixed>> input)

Apply this neural network in an input vector.