001/* 002 * Licensed to the Apache Software Foundation (ASF) under one or more 003 * contributor license agreements. See the NOTICE file distributed with 004 * this work for additional information regarding copyright ownership. 005 * The ASF licenses this file to You under the Apache License, Version 2.0 006 * (the "License"); you may not use this file except in compliance with 007 * the License. You may obtain a copy of the License at 008 * 009 * http://www.apache.org/licenses/LICENSE-2.0 010 * 011 * Unless required by applicable law or agreed to in writing, software 012 * distributed under the License is distributed on an "AS IS" BASIS, 013 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 014 * See the License for the specific language governing permissions and 015 * limitations under the License. 016 */ 017 018package org.apache.commons.math4.neuralnet.sofm.util; 019 020import java.util.function.DoubleUnaryOperator; 021import java.util.function.LongToDoubleFunction; 022 023import org.apache.commons.math4.neuralnet.internal.NeuralNetException; 024 025/** 026 * Decay function whose shape is similar to a sigmoid. 027 * <br> 028 * Class is immutable. 029 * 030 * @since 3.3 031 */ 032public class QuasiSigmoidDecayFunction implements LongToDoubleFunction { 033 /** Sigmoid. */ 034 private final DoubleUnaryOperator sigmoid; 035 /** See {@link #value(long)}. */ 036 private final double scale; 037 038 /** 039 * Creates an instance. 040 * The function {@code f} will have the following properties: 041 * <ul> 042 * <li>{@code f(0) = initValue}</li> 043 * <li>{@code numCall} is the inflexion point</li> 044 * <li>{@code slope = f'(numCall)}</li> 045 * </ul> 046 * 047 * @param initValue Initial value, i.e. {@link #applyAsDouble(long) applyAsDouble(0)}. 048 * @param slope Value of the function derivative at {@code numCall}. 049 * @param numCall Inflexion point. 050 * @throws IllegalArgumentException if {@code initValue <= 0}, 051 * {@code slope >= 0} or {@code numCall <= 0}. 052 */ 053 public QuasiSigmoidDecayFunction(double initValue, 054 double slope, 055 long numCall) { 056 if (initValue <= 0) { 057 throw new NeuralNetException(NeuralNetException.NOT_STRICTLY_POSITIVE, initValue); 058 } 059 if (slope >= 0) { 060 throw new NeuralNetException(NeuralNetException.TOO_LARGE, slope, 0); 061 } 062 if (numCall <= 1) { 063 throw new NeuralNetException(NeuralNetException.TOO_SMALL, numCall, 1); 064 } 065 066 final double k = initValue; 067 final double m = numCall; 068 final double b = 4 * slope / initValue; 069 sigmoid = x -> k / (1 + Math.exp(b * (m - x))); 070 071 final double y0 = sigmoid.applyAsDouble(0d); 072 scale = k / y0; 073 } 074 075 /** 076 * Computes the value of the learning factor. 077 * 078 * @param numCall Current step of the training task. 079 * @return the value of the function at {@code numCall}. 080 */ 081 @Override 082 public double applyAsDouble(long numCall) { 083 return scale * sigmoid.applyAsDouble((double) numCall); 084 } 085}