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 
18 package org.apache.commons.math3.optimization.general;
19 
20 /**
21  * This interface represents a preconditioner for differentiable scalar
22  * objective function optimizers.
23  * @deprecated As of 3.1 (to be removed in 4.0).
24  * @since 2.0
25  */
26 @Deprecated
27 public interface Preconditioner {
28     /**
29      * Precondition a search direction.
30      * <p>
31      * The returned preconditioned search direction must be computed fast or
32      * the algorithm performances will drop drastically. A classical approach
33      * is to compute only the diagonal elements of the hessian and to divide
34      * the raw search direction by these elements if they are all positive.
35      * If at least one of them is negative, it is safer to return a clone of
36      * the raw search direction as if the hessian was the identity matrix. The
37      * rationale for this simplified choice is that a negative diagonal element
38      * means the current point is far from the optimum and preconditioning will
39      * not be efficient anyway in this case.
40      * </p>
41      * @param point current point at which the search direction was computed
42      * @param r raw search direction (i.e. opposite of the gradient)
43      * @return approximation of H<sup>-1</sup>r where H is the objective function hessian
44      */
precondition(double[] point, double[] r)45     double[] precondition(double[] point, double[] r);
46 }
47