Package com.jml.linear_models

Class MultipleLinearRegressionSGD

java.lang.Object

com.jml.core.Model<double[][],double[]>

com.jml.linear_models.MultipleLinearRegression

com.jml.linear_models.MultipleLinearRegressionSGD

public class MultipleLinearRegressionSGD
extends MultipleLinearRegression

Model for least squares linear regression of multiple variables by stochastic gradient descent.

MultipleLinearRegressionSGD fits a model y = b₀ + b₁x₁ + ... + b_nx_n to the datasets by minimizing the residuals of the sum of squares (i.e. the sum of square errors) between the values in the target dataset and the values predicted by the model. This is minimized using stochastic gradient descent.

Constructor Summary

Constructors

Constructor	Description
MultipleLinearRegressionSGD()	Creates a MultipleLinearRegressionSGD model.
MultipleLinearRegressionSGD(double learningRate)	Creates a MultipleLinearRegressionSGD model.
MultipleLinearRegressionSGD(double learningRate, int maxIterations)	Creates a MultipleLinearRegressionSGD model.
MultipleLinearRegressionSGD(double learningRate, int maxIterations, double threshold)	Creates a MultipleLinearRegressionSGD model.
MultipleLinearRegressionSGD(int maxIterations)	Creates a MultipleLinearRegressionSGD model.

Method Summary

Modifier and Type	Method	Description
MultipleLinearRegressionSGD	fit(double[][] features, double[] targets)	Fits or trains the model with the given features and targets.
double[]	getLossHist()	Gets the loss history from the optimizer.

Methods inherited from class com.jml.linear_models.MultipleLinearRegression

getParams, inspect, predict, predict, saveModel, toString

Methods inherited from class com.jml.core.Model

load

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Details

MultipleLinearRegressionSGD

public MultipleLinearRegressionSGD()

Creates a MultipleLinearRegressionSGD model. This will use a default learning rate of 0.002.

MultipleLinearRegressionSGD

public MultipleLinearRegressionSGD(double learningRate,
 int maxIterations,
 double threshold)

Creates a MultipleLinearRegressionSGD model. When the fit method is called, Stochastic Gradient Descent will use the provided learning rate and will stop if it does not converge within the threshold by the specified number of max iterations.

Parameters:

learningRate - Learning rate to use during Stochastic Gradient Descent

threshold - Threshold for early stopping during Stochastic Gradient Descent. If the loss is less than the specified threshold, gradient descent will stop early.

maxIterations - Maximum number of iterations to run for during Stochastic Gradient Descent.

MultipleLinearRegressionSGD

public MultipleLinearRegressionSGD(double learningRate,
 int maxIterations)

Creates a MultipleLinearRegressionSGD model. When the fit method is called, Stochastic Gradient Descent will use the provided learning rate and will stop if it does not converge by the specified number of max iterations.

Parameters:

learningRate - Learning rate to use during Stochastic Gradient Descent.

maxIterations - Maximum number of iterations to run for during Stochastic Gradient Descent.

MultipleLinearRegressionSGD

public MultipleLinearRegressionSGD(double learningRate)

Creates a MultipleLinearRegressionSGD model. When the fit method is called, Stochastic Gradient Descent will use the provided learning rate and will stop if it does not converge by the specified number of max iterations.

Parameters:

learningRate - Learning rate to use during Stochastic Gradient Descent.

MultipleLinearRegressionSGD

public MultipleLinearRegressionSGD(int maxIterations)

Creates a MultipleLinearRegressionSGD model. When the fit method is called, Stochastic Gradient Descent will use the provided learning rate and will stop if it does not converge by the specified number of max iterations.

Parameters:

maxIterations - Maximum number of iterations to run for during Stochastic Gradient Descent.

Method Details

fit

public MultipleLinearRegressionSGD fit(double[][] features,
 double[] targets)

Fits or trains the model with the given features and targets. For both the features and targets parameters, if they are 2D arrays, then the number of rows in each must match and will be the number of samples in the data. The number of columns in each will be the number of features and targets in a single sample.

For instance, if the features array has shape n-by-m and the targets array has shape n-by-k. Then there are n samples in the dataset, each individual sample has m features, and each individual sample has k targets.

Overrides:

fit in class MultipleLinearRegression

Parameters:

features - The features of the training set.

targets - The targets of the training set.

Returns:

This. i.e. the trained model.

getLossHist

public double[] getLossHist()

Gets the loss history from the optimizer.

Returns:

Returns the loss for each iteration of the optimization algorithm in an array. The index of the array corresponds to the iteration the loss was computed for.