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.math4.neuralnet.sofm.util; 19 20 import java.util.function.LongToDoubleFunction; 21 22 import org.apache.commons.math4.neuralnet.internal.NeuralNetException; 23 24 /** 25 * Exponential decay function: <code>a e<sup>-x / b</sup></code>, 26 * where {@code x} is the (integer) independent variable. 27 * <br> 28 * Class is immutable. 29 * 30 * @since 3.3 31 */ 32 public class ExponentialDecayFunction implements LongToDoubleFunction { 33 /** Factor {@code a}. */ 34 private final double a; 35 /** Factor {@code 1 / b}. */ 36 private final double oneOverB; 37 38 /** 39 * Creates an instance. It will be such that 40 * <ul> 41 * <li>{@code a = initValue}</li> 42 * <li>{@code b = -numCall / ln(valueAtNumCall / initValue)}</li> 43 * </ul> 44 * 45 * @param initValue Initial value, i.e. {@link #applyAsDouble(long) applyAsDouble(0)}. 46 * @param valueAtNumCall Value of the function at {@code numCall}. 47 * @param numCall Argument for which the function returns 48 * {@code valueAtNumCall}. 49 * @throws IllegalArgumentException if {@code initValue <= 0}, 50 * {@code valueAtNumCall <= 0}, {@code valueAtNumCall >= initValue} or 51 * {@code numCall <= 0}. 52 */ 53 public ExponentialDecayFunction(double initValue, 54 double valueAtNumCall, 55 long numCall) { 56 if (initValue <= 0) { 57 throw new NeuralNetException(NeuralNetException.NOT_STRICTLY_POSITIVE, initValue); 58 } 59 if (valueAtNumCall <= 0) { 60 throw new NeuralNetException(NeuralNetException.NOT_STRICTLY_POSITIVE, valueAtNumCall); 61 } 62 if (valueAtNumCall >= initValue) { 63 throw new NeuralNetException(NeuralNetException.TOO_LARGE, valueAtNumCall, initValue); 64 } 65 if (numCall <= 0) { 66 throw new NeuralNetException(NeuralNetException.NOT_STRICTLY_POSITIVE, numCall); 67 } 68 69 a = initValue; 70 oneOverB = -Math.log(valueAtNumCall / initValue) / numCall; 71 } 72 73 /** 74 * Computes <code>a e<sup>-numCall / b</sup></code>. 75 * 76 * @param numCall Current step of the training task. 77 * @return the value of the function at {@code numCall}. 78 */ 79 @Override 80 public double applyAsDouble(long numCall) { 81 return a * Math.exp(-numCall * oneOverB); 82 } 83 }