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; 019 020import org.apache.commons.math4.neuralnet.internal.NeuralNetException; 021import org.apache.commons.math4.neuralnet.sofm.util.ExponentialDecayFunction; 022import org.apache.commons.math4.neuralnet.sofm.util.QuasiSigmoidDecayFunction; 023 024/** 025 * Factory for creating instances of {@link LearningFactorFunction}. 026 * 027 * @since 3.3 028 */ 029public final class LearningFactorFunctionFactory { 030 /** Class contains only static methods. */ 031 private LearningFactorFunctionFactory() {} 032 033 /** 034 * Creates an exponential decay {@link LearningFactorFunction function}. 035 * It will compute <code>a e<sup>-x / b</sup></code>, 036 * where {@code x} is the (integer) independent variable and 037 * <ul> 038 * <li><code>a = initValue</code> 039 * <li><code>b = -numCall / ln(valueAtNumCall / initValue)</code> 040 * </ul> 041 * 042 * @param initValue Initial value, i.e. 043 * {@link LearningFactorFunction#value(long) value(0)}. 044 * @param valueAtNumCall Value of the function at {@code numCall}. 045 * @param numCall Argument for which the function returns 046 * {@code valueAtNumCall}. 047 * @return the learning factor function. 048 * @throws IllegalArgumentException if {@code initValue <= 0}, 049 * {@code initValue > 1} {@code valueAtNumCall <= 0}, 050 * {@code valueAtNumCall >= initValue} or {@code numCall <= 0}. 051 */ 052 public static LearningFactorFunction exponentialDecay(final double initValue, 053 final double valueAtNumCall, 054 final long numCall) { 055 if (initValue <= 0 || 056 initValue > 1) { 057 throw new NeuralNetException(NeuralNetException.OUT_OF_RANGE, initValue, 0, 1); 058 } 059 060 return new LearningFactorFunction() { 061 /** DecayFunction. */ 062 private final ExponentialDecayFunction decay 063 = new ExponentialDecayFunction(initValue, valueAtNumCall, numCall); 064 065 /** {@inheritDoc} */ 066 @Override 067 public double value(long n) { 068 return decay.applyAsDouble(n); 069 } 070 }; 071 } 072 073 /** 074 * Creates an sigmoid-like {@code LearningFactorFunction function}. 075 * The function {@code f} will have the following properties: 076 * <ul> 077 * <li>{@code f(0) = initValue}</li> 078 * <li>{@code numCall} is the inflexion point</li> 079 * <li>{@code slope = f'(numCall)}</li> 080 * </ul> 081 * 082 * @param initValue Initial value, i.e. 083 * {@link LearningFactorFunction#value(long) value(0)}. 084 * @param slope Value of the function derivative at {@code numCall}. 085 * @param numCall Inflexion point. 086 * @return the learning factor function. 087 * @throws IllegalArgumentException if {@code initValue <= 0}, 088 * {@code initValue > 1}, {@code slope >= 0} or {@code numCall <= 0}. 089 */ 090 public static LearningFactorFunction quasiSigmoidDecay(final double initValue, 091 final double slope, 092 final long numCall) { 093 if (initValue <= 0 || 094 initValue > 1) { 095 throw new NeuralNetException(NeuralNetException.OUT_OF_RANGE, initValue, 0, 1); 096 } 097 098 return new LearningFactorFunction() { 099 /** DecayFunction. */ 100 private final QuasiSigmoidDecayFunction decay 101 = new QuasiSigmoidDecayFunction(initValue, slope, numCall); 102 103 /** {@inheritDoc} */ 104 @Override 105 public double value(long n) { 106 return decay.applyAsDouble(n); 107 } 108 }; 109 } 110}