de.jstacs.sequenceScores.statisticalModels.trainable.discrete.inhomogeneous
Class MEMTools

java.lang.Object
  de.jstacs.sequenceScores.statisticalModels.trainable.discrete.inhomogeneous.MEMTools

public class MEMTools
extends Object

Author:

Jens Keilwagen

Nested Class Summary

static class

MEMTools.DualFunction The dual function to the constraint problem of learning MEM's.

Field Summary

static byte	BGIS This constant can be used to specify that the model should use the blockwise iterative scaling for training.
static byte	BGIS_P This constant can be used to specify that the model should use the blockwise iterative scaling for training.
static byte	GIS This constant can be used to specify that the model should use the iterative scaling for training.
protected static double	LIN_EPS The epsilon for the line search in an optimization using the Optimizer.
static byte	SGIS This constant can be used to specify that the model should use the iterative scaling for training.
static byte	SGIS_P This constant can be used to specify that the model should use the iterative scaling for training.
protected static double	STARTDISTANCE The start distance for the line search in an optimization using the Optimizer.

Constructor Summary

MEMTools()

Method Summary

static double	getExpPartOfProb(MEMConstraint[] constraints, int[] fulfilled, SequenceIterator sequence) This method computes the exponential part of the probability, i.e., everything except the normalization constant.
static void	setParametersToValue(MEMConstraint[] constraint, double val) This method is a convenience method that sets the same value for all parameter of the constraints
static double	train(MEMConstraint[] constraints, int[][] cond, SequenceIterator sequence, byte algorithm, TerminationCondition condition, OutputStream stream, int[] alphLen) This method approximates the distribution either analytically or numerically.

Methods inherited from class java.lang.Object

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

Field Detail

SGIS_P

public static final byte SGIS_P

This constant can be used to specify that the model should use the iterative scaling for training.

See Also:

Constant Field Values

BGIS_P

public static final byte BGIS_P

This constant can be used to specify that the model should use the blockwise iterative scaling for training.

See Also:

Constant Field Values

GIS

public static final byte GIS

This constant can be used to specify that the model should use the iterative scaling for training.

See Also:

Constant Field Values

SGIS

public static final byte SGIS

This constant can be used to specify that the model should use the iterative scaling for training.

See Also:

Constant Field Values

BGIS

public static final byte BGIS

This constant can be used to specify that the model should use the blockwise iterative scaling for training.

See Also:

Constant Field Values

LIN_EPS

protected static final double LIN_EPS

The epsilon for the line search in an optimization using the Optimizer.

See Also:

Constant Field Values

STARTDISTANCE

protected static final double STARTDISTANCE

The start distance for the line search in an optimization using the Optimizer.

See Also:

Constant Field Values

Constructor Detail

MEMTools

public MEMTools()

Method Detail

train

public static double train(MEMConstraint[] constraints,
                           int[][] cond,
                           SequenceIterator sequence,
                           byte algorithm,
                           TerminationCondition condition,
                           OutputStream stream,
                           int[] alphLen)
                    throws Exception

This method approximates the distribution either analytically or numerically.

Parameters:

constraints - the constraints to be used

cond - the conditions

sequence - the SequenceIterator used in normalization and numerical approximation

algorithm - the choice of numerical approximation

condition - the TerminationCondition for stopping the iterative algorithm

stream - a possibility for writing some information

alphLen - the alphabet length of all positions

Returns:

the normalization constant

Throws:

Exception - if something went wrong inside the algorithms

setParametersToValue

public static void setParametersToValue(MEMConstraint[] constraint,
                                        double val)

This method is a convenience method that sets the same value for all parameter of the constraints

Parameters:

constraint - the constraints

val - the value to be set

getExpPartOfProb

public static double getExpPartOfProb(MEMConstraint[] constraints,
                                      int[] fulfilled,
                                      SequenceIterator sequence)

This method computes the exponential part of the probability, i.e., everything except the normalization constant.

Parameters:

constraints - the constraint

fulfilled - an array allowing to store which specific constraint is used, can be null

sequence - a sequence iteration

Returns:

the exponential part of the probability