/*
* Encog(tm) Core v3.4 - Java Version
* http://www.heatonresearch.com/encog/
* https://github.com/encog/encog-java-core
* Copyright 2008-2016 Heaton Research, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
* For more information on Heaton Research copyrights, licenses
* and trademarks visit:
* http://www.heatonresearch.com/copyright
*/
package org.encog.mathutil.error;
import org.encog.Encog;
/**
* Calculate the error of a neural network. Encog currently supports three error
* calculation modes. See ErrorCalculationMode for more info.
*/
public class ErrorCalculation {
/**
* The current error calculation mode.
*/
private static ErrorCalculationMode mode = ErrorCalculationMode.MSE;
/**
* get the error calculation mode, this is static and therefore global to
* all Enocg training. If a particular training method only supports a
* particular error calculation method, it may override this value. It will
* not change the value set here, rather the training will occur with its
* preferred training method. Currently the only training method that does
* this is Levenberg Marquardt (LMA).
*
* The default error mode for Encog is MSE.
*
* @return The current mode.
*/
public static ErrorCalculationMode getMode() {
return ErrorCalculation.mode;
}
/**
* Set the error calculation mode, this is static and therefore global to
* all Enocg training. If a particular training method only supports a
* particular error calculation method, it may override this value. It will
* not change the value set here, rather the training will occur with its
* preferred training method. Currently the only training method that does
* this is Levenberg Marquardt (LMA).
*
* @param theMode
* The new mode.
*/
public static void setMode(final ErrorCalculationMode theMode) {
ErrorCalculation.mode = theMode;
}
/**
* The overall error.
*/
private double globalError;
/**
* The size of a set.
*/
private int setSize;
private double sum;
private double min;
private double max;
/**
* Returns the root mean square error for a complete training set.
*
* @return The current error for the neural network.
*/
public final double calculate() {
if (this.setSize == 0) {
return 0;
}
switch (ErrorCalculation.getMode()) {
case RMS:
return calculateRMS();
case MSE:
return calculateMSE();
case ESS:
return calculateESS();
case LOGLOSS:
case HOT_LOGLOSS:
return calculateLogLoss();
case NRMSE_MEAN:
return calculateMeanNRMSE();
case NRMSE_RANGE:
return calculateRangeNRMSE();
default:
return calculateMSE();
}
}
/**
* Calculate the error with MSE.
*
* @return The current error for the neural network.
*/
public final double calculateMSE() {
if (this.setSize == 0) {
return 0;
}
final double err = this.globalError / this.setSize;
return err;
}
/**
* Calculate the error with SSE.
*
* @return The current error for the neural network.
*/
public final double calculateESS() {
if (this.setSize == 0) {
return 0;
}
final double err = this.globalError / 2;
return err;
}
public final double calculateMeanNRMSE() {
return calculateRMS()/(this.sum/this.setSize);
}
public final double calculateRangeNRMSE() {
return calculateRMS()/(this.max-this.min);
}
/**
* Calculate the error with RMS.
*
* @return The current error for the neural network.
*/
public final double calculateRMS() {
if (this.setSize == 0) {
return 0;
}
final double err = Math.sqrt(this.globalError / this.setSize);
return err;
}
public final double calculateLogLoss() {
return this.globalError*(-1.0/this.setSize);
}
/**
* Reset the error accumulation to zero.
*/
public final void reset() {
this.globalError = 0;
this.setSize = 0;
}
/**
* Update the error with single values.
*
* @param actual
* The actual value.
* @param ideal
* The ideal value.
*/
public final void updateError(final double actual, final double ideal) {
if(ErrorCalculation.getMode()==ErrorCalculationMode.LOGLOSS || ErrorCalculation.getMode()==ErrorCalculationMode.HOT_LOGLOSS ) {
this.globalError += Math.log(actual) * ideal;
this.setSize++;
} else {
double delta = ideal - actual;
this.globalError += delta * delta;
this.sum+=ideal;
if( this.setSize==0) {
this.min = this.max = actual;
} else {
this.min = Math.min(actual,this.min);
this.max = Math.max(actual,this.max);
}
this.setSize++;
}
}
/**
* Called to update for each number that should be checked.
*
* @param actual
* The actual number.
* @param ideal
* The ideal number.
* @param significance The signficance.
*/
public final void updateError(final double[] actual, final double[] ideal, final double significance) {
if (ErrorCalculation.getMode()==ErrorCalculationMode.HOT_LOGLOSS) {
this.setSize++;
for (int i = 0; i < actual.length; i++) {
// Only do the log if needed (for performance)
if( ideal[i]> Encog.DEFAULT_DOUBLE_EQUAL ) {
this.globalError += Math.log(actual[i]) * ideal[i];
}
}
} else if (ErrorCalculation.getMode()==ErrorCalculationMode.LOGLOSS) {
this.setSize++;
this.globalError += Math.log(actual[(int)ideal[0]]);
} else {
for (int i = 0; i < actual.length; i++) {
double delta = (ideal[i] - actual[i]) * significance;
// Do not apply significance to sum, min, max, they are only used for normalized RMSE.
this.sum+=ideal[i];
if( this.setSize==0) {
this.min = this.max = actual[i];
} else {
this.min = Math.min(actual[i],this.min);
this.max = Math.max(actual[i],this.max);
}
this.globalError += delta * delta;
}
this.setSize += ideal.length;
}
}
}