org.apache.ignite.ml.math.isolve.lsqr

Class AbstractLSQR

java.lang.Object

org.apache.ignite.ml.math.isolve.lsqr.AbstractLSQR

Direct Known Subclasses:

LSQROnHeap

public abstract class AbstractLSQR
extends Object

Basic implementation of the LSQR algorithm without assumptions about dataset storage format or data processing device. This implementation is based on SciPy implementation. SciPy implementation: https://github.com/scipy/scipy/blob/master/scipy/sparse/linalg/isolve/lsqr.py#L98.

Constructor Summary

Constructors

Constructor and Description
AbstractLSQR()

Method Summary

Modifier and Type	Method and Description
protected abstract double	beta(double[] x, double alfa, double beta) Calculates beta.
protected abstract double	bnorm() Calculates bnorm.
protected abstract Integer	getColumns()
protected abstract double[]	iter(double bnorm, double[] target) Perform LSQR iteration.
LSQRResult	solve(double damp, double atol, double btol, double conlim, double iterLim, boolean calcVar, double[] x0) Solves given Sparse Linear Systems.

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractLSQR

public AbstractLSQR()

Method Detail

solve

public LSQRResult solve(double damp,
                        double atol,
                        double btol,
                        double conlim,
                        double iterLim,
                        boolean calcVar,
                        double[] x0)

Solves given Sparse Linear Systems.

Parameters:

damp - Damping coefficient.

atol - Stopping tolerances, if both (atol and btol) are 1.0e-9 (say), the final residual norm should be accurate to about 9 digits.

btol - Stopping tolerances, if both (atol and btol) are 1.0e-9 (say), the final residual norm should be accurate to about 9 digits.

conlim - Another stopping tolerance, LSQR terminates if an estimate of cond(A) exceeds conlim.

iterLim - Explicit limitation on number of iterations (for safety).

calcVar - Whether to estimate diagonals of (A'A + damp^2*I)^{-1}.

x0 - Initial value of x.

Returns:

Solver result.

bnorm

protected abstract double bnorm()

Calculates bnorm.

Returns:

bnorm

beta

protected abstract double beta(double[] x,
                               double alfa,
                               double beta)

Calculates beta.

Parameters:

x - X value.

alfa - Alfa value.

beta - Beta value.

Returns:

Beta.

iter

protected abstract double[] iter(double bnorm,
                                 double[] target)

Perform LSQR iteration.

Parameters:

bnorm - Bnorm value.

target - Target value.

Returns:

Iteration result.

getColumns

protected abstract Integer getColumns()