org.akutan.optimization
Interface LMObjective

public interface LMObjective

This is the interface implemented by a cost function suitable for optimization using the Levenberg-Marquardt algorithm. Note this is for a generic minimization, not a least squares fit.

Since:

10 may 2007

Method Summary

cern.colt.matrix.DoubleMatrix1D	gradient(cern.colt.matrix.DoubleMatrix1D x) Called to compute the gradient of the function at x, for each member of x.
cern.colt.matrix.DoubleMatrix2D	hessian(cern.colt.matrix.DoubleMatrix1D x) Computes an exact or approximated Hessian of the function at x.
double	value(cern.colt.matrix.DoubleMatrix1D x) Called to compute the value of the function at x

Method Detail

value

double value(cern.colt.matrix.DoubleMatrix1D x)

Called to compute the value of the function at x

Parameters:

x - The input vector to use in the calculation

Returns:

The cost value associated with the input vector.

gradient

cern.colt.matrix.DoubleMatrix1D gradient(cern.colt.matrix.DoubleMatrix1D x)

Called to compute the gradient of the function at x, for each member of x.

Parameters:

x - The input vector to use in the calculation.

Returns:

The gradient vector for the input point.

hessian

cern.colt.matrix.DoubleMatrix2D hessian(cern.colt.matrix.DoubleMatrix1D x)

Computes an exact or approximated Hessian of the function at x.

Parameters:

x - The input vector to use in the calculation

Returns:

The Hessian matrix for the input point.