iitb.CRF
Class CrfParams

java.lang.Object
  extended byiitb.CRF.CrfParams

public class CrfParams
extends java.lang.Object

This class holds all parameters to control various aspects of the CRF model

Author:
Sunita Sarawagi

Field Summary
 int debugLvl
           
 boolean doScaling
           
 double epsForConvergence
          Convergence criteria for finding optimum lambda using BFGS
 double initValue
          initial value for all the lambda arrays
 double invSigmaSquare
          penalty term for likelihood function is ||lambda||^2*invSigmaSquare/2 set this to zero, if no penalty needed
 int maxIters
          Maximum number of iterations over the training data during training
 int mForHessian
          The number of corrections used in the BFGS update.
 java.lang.String trainerType
           
 
Constructor Summary
CrfParams()
          constructor with default values as follows.
CrfParams(java.util.Properties opts)
           
CrfParams(java.lang.String args)
          Initialize any parameter using space separated list of name value pairs Example: "initValue 0.1 maxIters 20" will set the initValue param to 0.1 and maxIters param to 20.
 
Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Field Detail

initValue

public double initValue
initial value for all the lambda arrays


invSigmaSquare

public double invSigmaSquare
penalty term for likelihood function is ||lambda||^2*invSigmaSquare/2 set this to zero, if no penalty needed


maxIters

public int maxIters
Maximum number of iterations over the training data during training


epsForConvergence

public double epsForConvergence
Convergence criteria for finding optimum lambda using BFGS


mForHessian

public int mForHessian
The number of corrections used in the BFGS update.


trainerType

public java.lang.String trainerType

debugLvl

public int debugLvl

doScaling

public boolean doScaling
Constructor Detail

CrfParams

public CrfParams()
constructor with default values as follows. initValue = 0; invSigmaSquare = 1.0; maxIters = 50; epsForConvergence = 0.0001; mForHessian = 7; trainerType = ""; debugLvl = 1;


CrfParams

public CrfParams(java.lang.String args)
Initialize any parameter using space separated list of name value pairs Example: "initValue 0.1 maxIters 20" will set the initValue param to 0.1 and maxIters param to 20.


CrfParams

public CrfParams(java.util.Properties opts)


Submit a bug or feature.