Package dk.alexandra.fresco.stat

Class DefaultMachineLearning

Object

DefaultMachineLearning

All Implemented Interfaces:

MachineLearning

public class DefaultMachineLearning
extends Object
implements MachineLearning

Method Summary

Modifier and Type	Method	Description
DRes<MLP>	fit(MLP network, List<ArrayList<DRes<SFixed>>> data, List<ArrayList<DRes<SFixed>>> labels, int epochs, double learningRate)	Fit the given multilayer perceptron to a dataset using back propagation.
DRes<ArrayList<DRes<SFixed>>>	logisticRegression(Matrix<DRes<SFixed>> data, ArrayList<DRes<SFixed>> expected, double[] beta, IntToDoubleFunction rate, int epochs)	Estimate the parameters of a logistic model using gradient descent.
DRes<SInt>	predict(MLP network, ArrayList<DRes<SFixed>> input)	Assuming that the given MLP has n output neurons, this function applies the network to the given input and finds the index of the output i with 0 ≤ i < n containing the largest number.

Methods inherited from class Object

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

Method Details

logisticRegression

public DRes<ArrayList<DRes<SFixed>>> logisticRegression(Matrix<DRes<SFixed>> data, ArrayList<DRes<SFixed>> expected, double[] beta, IntToDoubleFunction rate, int epochs)

Description copied from interface: MachineLearning

Estimate the parameters of a logistic model using gradient descent.

Specified by:

logisticRegression in interface MachineLearning

Parameters:

data - The data represented as a matrix with entry as rows.

expected - The expected outcome for each entry represented as a list. Each entry should be either 0 or 1.

beta - The initial guess for the parameters of the model with the first being the constant term.

rate - The learning rate used by the gradient descent algorithm as a function of the iteration number.

epochs - The number of iterations.

Returns:

An approximation of the parameters of a logistic model fitting the given data.

fit

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

Description copied from interface: MachineLearning

Fit the given multilayer perceptron to a dataset using back propagation.

Specified by:

fit in interface MachineLearning

Parameters:

network - The mlp to fit.

data - The dataset to use as input.

labels - The expected outputs.

epochs - The number of epochs, ie. iterations through the entire dataset.

learningRate - The learning rate.

Returns:

A new MLP with updated weights.

predict

public DRes<SInt> predict(MLP network, ArrayList<DRes<SFixed>> input)

Description copied from interface: MachineLearning

Assuming that the given MLP has n output neurons, this function applies the network to the given input and finds the index of the output i with 0 ≤ i < n containing the largest number.

Specified by:

predict in interface MachineLearning

Parameters:

network - The multi-layer perceptron to use for the prediction.

input - The input.

Returns:

The index of the largest element in the output vector.