From 007a701755f318562b0cbf2005d84299733d9b45 Mon Sep 17 00:00:00 2001 From: Luc Maisonobe Date: Sun, 2 Mar 2014 19:54:43 +0000 Subject: [PATCH] Add SVD to GaussNewtonOptimizer Allow using the SVD decomposition which has excellent stability and can provide a "solution" for singular problems. Figured out why one of the least squares tests did not check two of the states. There was only one equation for the two states. The test now verifies the states satisfy the constraint. Patch provided by Evan Ward. JIRA: MATH-1104 git-svn-id: https://svn.apache.org/repos/asf/commons/proper/math/trunk@1573351 13f79535-47bb-0310-9956-ffa450edef68 --- .../leastsquares/GaussNewtonOptimizer.java | 26 ++- ...ractLeastSquaresOptimizerAbstractTest.java | 8 +- .../GaussNewtonOptimizerWithSVDTest.java | 149 ++++++++++++++++++ 3 files changed, 178 insertions(+), 5 deletions(-) create mode 100644 src/test/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizerWithSVDTest.java diff --git a/src/main/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizer.java b/src/main/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizer.java index c17c870c4..cdb48b6b3 100644 --- a/src/main/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizer.java +++ b/src/main/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizer.java @@ -29,15 +29,19 @@ import org.apache.commons.math3.linear.QRDecomposition; import org.apache.commons.math3.linear.RealMatrix; import org.apache.commons.math3.linear.RealVector; import org.apache.commons.math3.linear.SingularMatrixException; +import org.apache.commons.math3.linear.SingularValueDecomposition; import org.apache.commons.math3.optim.ConvergenceChecker; import org.apache.commons.math3.util.Incrementor; import org.apache.commons.math3.util.Pair; /** - * Gauss-Newton least-squares solver.

This class solve a least-square problem by + * Gauss-Newton least-squares solver. + *

This class solve a least-square problem by * solving the normal equations of the linearized problem at each iteration. Either LU - * decomposition or QR decomposition can be used to solve the normal equations. LU - * decomposition is faster but QR decomposition is more robust for difficult problems. + * decomposition or Cholesky decomposition can be used to solve the normal equations, + * or QR decomposition or SVD decomposition can be used to solve the linear system. LU + * decomposition is faster but QR decomposition is more robust for difficult problems, + * and SVD can compute a solution for rank-deficient problems. *

* * @version $Id$ @@ -119,6 +123,22 @@ public class GaussNewtonOptimizer implements LeastSquaresOptimizer { throw new ConvergenceException(LocalizedFormats.UNABLE_TO_SOLVE_SINGULAR_PROBLEM, e); } } + }, + /** + * Solve the linear least squares problem using the {@link + * SingularValueDecomposition}. + * + *

This method is slower, but can provide a solution for rank deficient and + * nearly singular systems. + */ + SVD { + @Override + protected RealVector solve(final RealMatrix jacobian, + final RealVector residuals) { + return new SingularValueDecomposition(jacobian) + .getSolver() + .solve(residuals); + } }; /** diff --git a/src/test/java/org/apache/commons/math3/fitting/leastsquares/AbstractLeastSquaresOptimizerAbstractTest.java b/src/test/java/org/apache/commons/math3/fitting/leastsquares/AbstractLeastSquaresOptimizerAbstractTest.java index 256925902..86c3ab37a 100644 --- a/src/test/java/org/apache/commons/math3/fitting/leastsquares/AbstractLeastSquaresOptimizerAbstractTest.java +++ b/src/test/java/org/apache/commons/math3/fitting/leastsquares/AbstractLeastSquaresOptimizerAbstractTest.java @@ -303,8 +303,12 @@ public abstract class AbstractLeastSquaresOptimizerAbstractTest { problem.getBuilder().start(new double[]{2, 2, 2, 2, 2, 2}).build()); Assert.assertEquals(0, optimum.getRMS(), TOl); - //TODO the first two elements of point were not previously checked - assertEquals(TOl, optimum.getPoint(), 2, 1, 3, 4, 5, 6); + RealVector point = optimum.getPoint(); + //the first two elements are under constrained + //check first two elements obey the constraint: sum to 3 + Assert.assertEquals(3, point.getEntry(0) + point.getEntry(1), TOl); + //#constrains = #states fro the last 4 elements + assertEquals(TOl, point.getSubVector(2, 4), 3, 4, 5, 6); } @Test diff --git a/src/test/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizerWithSVDTest.java b/src/test/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizerWithSVDTest.java new file mode 100644 index 000000000..a205d5398 --- /dev/null +++ b/src/test/java/org/apache/commons/math3/fitting/leastsquares/GaussNewtonOptimizerWithSVDTest.java @@ -0,0 +1,149 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.commons.math3.fitting.leastsquares; + +import org.apache.commons.math3.exception.ConvergenceException; +import org.apache.commons.math3.exception.TooManyEvaluationsException; +import org.apache.commons.math3.fitting.leastsquares.GaussNewtonOptimizer.Decomposition; +import org.apache.commons.math3.fitting.leastsquares.LeastSquaresOptimizer.Optimum; +import org.apache.commons.math3.geometry.euclidean.threed.Plane; +import org.apache.commons.math3.geometry.euclidean.threed.Vector3D; +import org.apache.commons.math3.optim.SimpleVectorValueChecker; +import org.apache.commons.math3.util.FastMath; +import org.junit.Assert; +import org.junit.Test; + +import java.io.IOException; + +/** + *

Some of the unit tests are re-implementations of the MINPACK file17 and file22 test files. + * The redistribution policy for MINPACK is available here/ + * + * @version $Id$ + */ +public class GaussNewtonOptimizerWithSVDTest + extends AbstractLeastSquaresOptimizerAbstractTest { + + @Override + public int getMaxIterations() { + return 1000; + } + + @Override + public LeastSquaresOptimizer getOptimizer() { + return new GaussNewtonOptimizer(Decomposition.SVD); + } + + @Test + public void testMaxEvaluations() throws Exception { + try{ + CircleVectorial circle = new CircleVectorial(); + circle.addPoint( 30.0, 68.0); + circle.addPoint( 50.0, -6.0); + circle.addPoint(110.0, -20.0); + circle.addPoint( 35.0, 15.0); + circle.addPoint( 45.0, 97.0); + + LeastSquaresProblem lsp = builder(circle) + .checkerPair(new SimpleVectorValueChecker(1e-30, 1e-30)) + .maxIterations(Integer.MAX_VALUE) + .start(new double[]{98.680, 47.345}) + .build(); + + optimizer.optimize(lsp); + + fail(optimizer); + }catch (TooManyEvaluationsException e){ + //expected + } + } + + @Override + @Test + public void testCircleFittingBadInit() { + /* + * This test converged to the wrong solution with this optimizer. + * It seems that the state becomes so large that the convergence + * checker's relative tolerance test passes. + */ + try { + super.testCircleFittingBadInit(); + fail(optimizer); + } catch (AssertionError e) { + //expected + } + } + + @Override + @Test + public void testHahn1() + throws IOException { + /* + * TODO This test leads to a singular problem with the Gauss-Newton + * optimizer. This should be inquired. + */ + try{ + super.testHahn1(); + fail(optimizer); + } catch (ConvergenceException e){ + //expected for LU + } catch (TooManyEvaluationsException e){ + //expected for QR + } + } + + @Test + @Override + public void testGetIterations() { + /* this diverges with SVD */ + try { + super.testGetIterations(); + fail(optimizer); + } catch (TooManyEvaluationsException e) { + //expected + } + } + + @Test + @Override + public void testNonInvertible() throws Exception { + /* SVD can compute a solution to singular problems. + * In this case the target vector, b, is not in the + * span of the jacobian matrix, A. The closes point + * to b on the plane spanned by A is computed. + */ + LinearProblem problem = new LinearProblem(new double[][]{ + {1, 2, -3}, + {2, 1, 3}, + {-3, 0, -9} + }, new double[]{1, 1, 1}); + + Optimum optimum = optimizer.optimize(problem.getBuilder().build()); + + Plane span = new Plane(Vector3D.ZERO, new Vector3D(1, 2, -3), new Vector3D(2, 1, 0), TOl); + double expected = FastMath.abs(span.getOffset(new Vector3D(1, 1, 1))); + double actual = optimum.getResiduals().getNorm(); + + //verify + Assert.assertEquals(expected, actual, TOl); + } + +}