Package jline.lib.perm

Class HeuristicPermanent

java.lang.Object

jline.lib.perm.PermSolver

jline.lib.perm.HeuristicPermanent

public class HeuristicPermanent
extends PermSolver

Heuristic approximation to the permanent of a positive matrix. This implementation uses Sinkhorn scaling to make the matrix approximately doubly stochastic, then applies a mean-field approximation with van der Waerden bounds combined with a Gurvits-like capacity bound. The algorithm: 1. Uses Sinkhorn scaling to make A approximately doubly stochastic 2. Applies a mean-field approximation with van der Waerden bounds 3. Combines with a Gurvits-like capacity bound 4. Scales the result back to the original matrix scale This is a heuristic approximation suitable for large matrices where exact computation is computationally prohibitive. The approximation quality depends on the structure of the input matrix.

Field Summary

Fields inherited from class jline.lib.perm.PermSolver

matrix, memory, n, time, value

Constructor Summary

Constructors

Constructor	Description
HeuristicPermanent(Matrix matrix)	Constructor with just matrix parameter.
HeuristicPermanent(Matrix matrix, boolean solve)	Constructor with just matrix and solve parameters.
HeuristicPermanent(Matrix matrix, double tolerance, int maxIterations, boolean solve)	Constructor with all parameters.

Method Summary

Modifier and Type	Method	Description
void	compute()	Compute the permanent or approximation for the given matrix.

Methods inherited from class jline.lib.perm.PermSolver

getMatrix, getMemory, getN, getTime, getValue, solve

Methods inherited from class java.lang.Object

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

Constructor Details

HeuristicPermanent

public HeuristicPermanent(Matrix matrix,
 double tolerance,
 int maxIterations,
 boolean solve)

Constructor with all parameters.

Parameters:

matrix - The matrix for which to compute the permanent (must be strictly positive)

tolerance - Convergence threshold for Sinkhorn scaling

maxIterations - Maximum number of Sinkhorn iterations

solve - Whether to automatically run solve() after construction

Throws:

IllegalArgumentException - if matrix contains non-positive elements

HeuristicPermanent

public HeuristicPermanent(Matrix matrix,
 boolean solve)

Constructor with just matrix and solve parameters. Uses default tolerance (1e-10) and maxIterations (1000).

HeuristicPermanent

public HeuristicPermanent(Matrix matrix)

Constructor with just matrix parameter. Uses default tolerance (1e-10), maxIterations (1000), and solve (false).

Method Details

compute

public void compute()

Description copied from class: PermSolver

Compute the permanent or approximation for the given matrix. This method must be implemented by all concrete solver classes. The result should be stored in the value field.

Specified by:

compute in class PermSolver