Package com.jml.linear_models

Class LinearRegressionSGD

java.lang.Object
com.jml.core.Model<double[],​double[]>
com.jml.linear_models.PolynomialRegression
com.jml.linear_models.LinearRegression
com.jml.linear_models.LinearRegressionSGD
public class LinearRegressionSGD extends LinearRegression
Model for least squares linear regression of one variable by stochastic gradient descent.

LinearRegressionSGD fits a model y = b0 + b1x to the datasets by minimizing the residuals of the sum of squares between the values in the target dataset and the values predicted by the model. This is using stochastic gradient descent.

  • Constructor Details

    • LinearRegressionSGD

      public LinearRegressionSGD()
      Creates a LinearRegressionSGD model.
      This will use default settings for gradient descent: 
          Learning Rate: 0.002
    Threshold: 0.5e-5
    Maximum Iterations: 1000
    Scheduler: None
 
      

      

      • 

    • 

      

      LinearRegressionSGD
      

      public LinearRegressionSGD​(double learningRate,
 int maxIterations,
 double threshold)
      

      Creates a LinearRegressionSGD 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.
      

      

      

      • 

    • 

      

      LinearRegressionSGD
      

      public LinearRegressionSGD​(double learningRate,
 int maxIterations)
      

      Creates a LinearRegressionSGD 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.
      

      

      

      • 

    • 

      

      LinearRegressionSGD
      

      public LinearRegressionSGD​(double learningRate)
      

      Creates a LinearRegressionSGD 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.
      

      

      

      • 

    • 

      

      LinearRegressionSGD
      

      public LinearRegressionSGD​(int maxIterations)
      

      Creates a LinearRegressionSGD 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 LinearRegressionSGD fit​(double[] features,
 double[] targets)
      

      Fits or trains the model with the given features and targets.
      

      

      Overrides:
      

      fit in class LinearRegression
      

      Parameters:
      

      features - The features of the training set.
      

      targets - The targets of the training set.
      

      Returns:
      

      Returns details of the fitting / training process.
      

      Throws:
      

      IllegalArgumentException - Can be thrown for the following reasons
      
                                  - If key, value pairs in args are unspecified or invalid arguments. 
      
                                  - If the features and targets are not correctly sized per the specification when the model was
                                  compiled.
      

      

      

      • 

    • 

      

      getLossHist
      

      public double[] getLossHist()
      

      Gets the loss history from training.
      

      

      Returns:
      

      The loss of every iteration stored in a List.
      

      

      

      • 

    

    

    •