1/*
2 * Licensed to the Apache Software Foundation (ASF) under one or more
3 * contributor license agreements.  See the NOTICE file distributed with
4 * this work for additional information regarding copyright ownership.
5 * The ASF licenses this file to You under the Apache License, Version 2.0
6 * (the "License"); you may not use this file except in compliance with
7 * the License.  You may obtain a copy of the License at
8 *
9 *      http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 */
17
18package org.apache.commons.math.optimization.general;
19
20import org.apache.commons.math.FunctionEvaluationException;
21
22/**
23 * This interface represents a preconditioner for differentiable scalar
24 * objective function optimizers.
25 * @version $Revision: 1073158 $ $Date: 2011-02-21 22:46:52 +0100 (lun. 21 févr. 2011) $
26 * @since 2.0
27 */
28public interface Preconditioner {
29
30    /**
31     * Precondition a search direction.
32     * <p>
33     * The returned preconditioned search direction must be computed fast or
34     * the algorithm performances will drop drastically. A classical approach
35     * is to compute only the diagonal elements of the hessian and to divide
36     * the raw search direction by these elements if they are all positive.
37     * If at least one of them is negative, it is safer to return a clone of
38     * the raw search direction as if the hessian was the identity matrix. The
39     * rationale for this simplified choice is that a negative diagonal element
40     * means the current point is far from the optimum and preconditioning will
41     * not be efficient anyway in this case.
42     * </p>
43     * @param point current point at which the search direction was computed
44     * @param r raw search direction (i.e. opposite of the gradient)
45     * @return approximation of H<sup>-1</sup>r where H is the objective function hessian
46     * @exception FunctionEvaluationException if no cost can be computed for the parameters
47     * @exception IllegalArgumentException if point dimension is wrong
48     */
49    double[] precondition(double[] point, double[] r)
50        throws FunctionEvaluationException, IllegalArgumentException;
51
52}
53