magpie.models.regression

Class AbstractNonlinearRegression

java.lang.Object

magpie.models.BaseModel

magpie.models.regression.BaseRegression

magpie.models.regression.AbstractNonlinearRegression

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, AbstractRegressionModel, Citable, Commandable, Options, Printable, Savable

Direct Known Subclasses:

AbstractParsedNonlinearRegression, SimpleLinearModelExample

public abstract class AbstractNonlinearRegression
extends BaseRegression

Abstract class for models that fit terms using non-linear regression. Implementations of this class must provide information about the function being fitted by fulfilling these operations.

1. function(double[], double[]) - Function to be fitted. Expressed as a function of independent variables and fitting coefficients.
2. defineVariables() - Map the variables in function to attributes.
3. defineCoefficients() - Define names of fitting coefficients

Usage: *No options*

A simple example of these operations is shown in the source code of SimpleLinearModelExample.

Currently, this code uses Powell's algorithm as implemented in Apache Common's Math Library: PowellOptimizer

Implemented Commands:

nonlinear guess <name> <guess> - Set the initial guess for a certain parameter
name: Name of parameter to define guess for
guess: Initial guess for that parameter

nonlinear maxiter <maxiter> - Set maximum number of iterations for fitting routines
maxiter: Maximum number of iterations

See Also:

Serialized Form

Field Summary

Fields inherited from class magpie.models.BaseModel

AttributeSelector, trained, TrainingStats, validated, ValidationStats

Constructor Summary

Constructors

Constructor and Description
AbstractNonlinearRegression()

Method Summary

Modifier and Type	Method and Description
protected void	addCoefficient(java.lang.String Name) Adds a new fitting parameter to function (if not already defined).
protected void	addCoefficient(java.lang.String Name, java.lang.Double Guess) Adds a new fitting parameter to function (if not already defined)
protected void	addVariable(java.lang.String Name) Define a variable to be used in the user function (if not already defined)
protected abstract void	defineCoefficients() Define the names and, if desired, initial guesses of fitting coefficients.
protected abstract void	defineVariables() Define which attributes are used by the function.
protected abstract double	function(double[] variables, double[] coeff) Function defined by the user.
java.lang.String	getCoefficientName(int index) Return name of coefficient, given index
double	getFittedCoefficient(int index) Get the value of fitted coefficient
int	getMaxIter() Get the maximum number of iterations allowed
int	getNFittingParameters() Number of fitting parameters in a model.
java.lang.String	getVariableName(int index)
protected org.apache.commons.math3.analysis.MultivariateFunction	makeObjectiveFunction(Dataset Data) Generate an anonymous MultivariateFunction object that calculates the error between the equation described by the user as a function of the fitting coefficients.
int	NCoefficients()
protected int	NVariables()
java.lang.String	printUsage() Print out required format for options.
void	run_protected(Dataset TrainData) Run a model without checking if stuff is trained (use carefully)
java.lang.Object	runCommand(java.util.List<java.lang.Object> Command) Process some command described by a list of Objects.
java.lang.Object	runNonlinearCommand(java.util.List<java.lang.Object> Command) Handle commands for nonliear models
void	setInitialGuess(java.lang.String Name, double value) Set the initial guess for a coefficient
void	setMaxIterations(int MaxIter) Define the maximum number of iterations allowed.
void	setOptions(java.util.List Options) Set any options for this object.
protected void	train_protected(Dataset TrainData) Train a model without evaluating performance

Methods inherited from class magpie.models.regression.BaseRegression

clone, doRobustRegression, getRobustRegressionQ, robustTraining, setRobustRegressionQ

Methods inherited from class magpie.models.BaseModel

about, crossValidate, crossValidate, crossValidate, done, externallyValidate, getAttributeSelector, getCitations, getFilter, getTrainTime, getValidationMethod, handleSetCommand, isTrained, isValidated, loadState, printCommand, printDescription, printModel_protected, printModel, printModelDescriptionDetails, resetModel, run, saveCommand, saveState, setAttributeSelector, setComponent, setFilter, train, train

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractNonlinearRegression

public AbstractNonlinearRegression()

Method Detail

setOptions

public void setOptions(java.util.List Options)
                throws java.lang.Exception

Description copied from interface: Options

Set any options for this object. If parsing the options fails, return an Exception with the usage as its message.

Parameters:

Options - Array of options as Objects - can be null

Throws:

java.lang.Exception - if problem with inputs

printUsage

public java.lang.String printUsage()

Description copied from interface: Options

Print out required format for options. For consistency, should start with "Usage: ".

Returns:

Usage for this function

setMaxIterations

public void setMaxIterations(int MaxIter)

Define the maximum number of iterations allowed. Anything less than 0 is interpreted as functionally infinite.

Parameters:

MaxIter - Desired maximum iteration count.

getMaxIter

public int getMaxIter()

Get the maximum number of iterations allowed

Returns:

Maximum # of iterations

getFittedCoefficient

public double getFittedCoefficient(int index)

Get the value of fitted coefficient

Parameters:

index - Index of coefficient to retrieve

Returns:

Value of that coefficient

makeObjectiveFunction

protected org.apache.commons.math3.analysis.MultivariateFunction makeObjectiveFunction(Dataset Data)

Generate an anonymous MultivariateFunction object that calculates the error between the equation described by the user as a function of the fitting coefficients.

Parameters:

Data - Dataset used to calculate the error

Returns:

Multivariate function as described above

defineVariables

protected abstract void defineVariables()

Define which attributes are used by the function. This function is run a single time during training. As long as variables are defined by the time this function is complete, it does not matter how your code actually does it.

Variables are defined using the addVariable(java.lang.String) function. *

In function(double[], double[]) these names are mapped to the variable array in the same order you define them.

addVariable

protected void addVariable(java.lang.String Name)

Define a variable to be used in the user function (if not already defined)

Parameters:

Name - Name of attribute to be include in equation.

NVariables

protected int NVariables()

Returns:

Number of variables used by the user-defined function

getVariableName

public java.lang.String getVariableName(int index)

Parameters:

index - Index of variable to be retrieved

Returns:

Name of that attribute variable

defineCoefficients

protected abstract void defineCoefficients()

Define the names and, if desired, initial guesses of fitting coefficients. As long as coefficients are defined by the time this program exits, it does not matter what this code actually does (i.e. you can define coefficients in other functions).

Coefficients are defined using the addCoefficient(java.lang.String) function.

In function(double[], double[]) these names are mapped to the coeff array in the same order you define them.

addCoefficient

protected void addCoefficient(java.lang.String Name)

Adds a new fitting parameter to function (if not already defined). Uses a default guess of 1.0.

Parameters:

Name - Name of fitting parameter

addCoefficient

protected void addCoefficient(java.lang.String Name,
                              java.lang.Double Guess)

Adds a new fitting parameter to function (if not already defined)

Parameters:

Name - Name of fitting parameter

Guess - Initial guess

NCoefficients

public int NCoefficients()

Returns:

Number of fitting parameters in function

getNFittingParameters

public int getNFittingParameters()

Description copied from interface: AbstractRegressionModel

Number of fitting parameters in a model. Only pertinent to nonlinear regression. Otherwise, this should just return 0.

Returns:

Number of fitting parameters

getCoefficientName

public java.lang.String getCoefficientName(int index)

Return name of coefficient, given index

Parameters:

index - Index of coefficient to

Returns:

Name of that coefficient

setInitialGuess

public void setInitialGuess(java.lang.String Name,
                            double value)
                     throws java.lang.Exception

Set the initial guess for a coefficient

Parameters:

Name - Name of coefficient to set

value - Desired initial guess

Throws:

java.lang.Exception - If there is no coefficient by that name

function

protected abstract double function(double[] variables,
                                   double[] coeff)

Function defined by the user. Program supplies attributes in the same order as defined in defineVariables().

Parameters:

variables - Values of attributes used as independent variables

coeff - Coefficients of equation

Returns:

Value of equation given those inputs

train_protected

protected void train_protected(Dataset TrainData)

Description copied from class: BaseModel

Train a model without evaluating performance

Specified by:

train_protected in class BaseModel

Parameters:

TrainData - Training data

run_protected

public void run_protected(Dataset TrainData)

Description copied from class: BaseModel

Run a model without checking if stuff is trained (use carefully)

Specified by:

run_protected in class BaseModel

Parameters:

TrainData - Training data

runCommand

public java.lang.Object runCommand(java.util.List<java.lang.Object> Command)
                            throws java.lang.Exception

Description copied from interface: Commandable

Process some command described by a list of Objects.

Specified by:

runCommand in interface Commandable

Overrides:

runCommand in class BaseRegression

Parameters:

Command - Command as a list of objects

Returns:

Any output (null if no output generated by command)

Throws:

java.lang.Exception - If something goes wrong

runNonlinearCommand

public java.lang.Object runNonlinearCommand(java.util.List<java.lang.Object> Command)
                                     throws java.lang.Exception

Handle commands for nonliear models

Parameters:

Command -

Returns:

Throws:

java.lang.Exception