com.javanut.pronghorn.util.ma.RunningStdDev Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of pronghorn-pipes Show documentation
Ring buffer based queuing utility for applications that require high performance and/or a small footprint. Well suited for embedded and stream based processing.
There is a newer version: 1.1.27
Show newest version
package com.javanut.pronghorn.util.ma;

import java.io.IOException;

import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import com.javanut.pronghorn.util.Appendables;

//Algo taken from knuth, "the art of computer programming"
public class RunningStdDev {
    final static Logger logger = LoggerFactory.getLogger(RunningStdDev.class);
   
    private long sampleCount;
    private double oldMean;
    private double newMean;
    private double oldStd;
    private double newStd;
    private double min;
    private double max;
    
    public void clear() {
            sampleCount = 0;
            oldMean = 0;
            newMean = 0;
            oldStd = 0;
            newStd = 0;
            min = Long.MAX_VALUE;
            max = Long.MIN_VALUE;
    }
    
    public RunningStdDev() {
        clear();
    }
    
    private static double probabilityDensityY(double x, double variance, double stdDev, double mean) {
        //computed http://www.had2know.com/academics/normal-distribution-probability-calculator.html
        double pow = -((x-mean)*(x-mean))/(2d*variance);
        double numerator = Math.pow(Math.E,pow);
        double denominator = stdDev*Math.sqrt(2d*Math.PI);
        double result = numerator/denominator;
        return result;
    }

    public static void sample(RunningStdDev runStdDev, double sample) {
            if (Double.isNaN(sample)) {
                return; //do not add sample if its not a number
            }

            if (samplerunStdDev.max) {
            	runStdDev.max = sample;
            }
            runStdDev.sampleCount++;
            // See Knuth TAOCP vol 2, 3rd edition, page 232
            if (runStdDev.sampleCount == 1)
            {
            	runStdDev.oldMean = runStdDev.newMean = sample;
            	runStdDev.oldStd = 0.0;
            } else {
            	double temp = sample - runStdDev.oldMean;
            	runStdDev.newMean = runStdDev.oldMean + temp/runStdDev.sampleCount;
            	runStdDev.newStd = runStdDev.oldStd + temp*(sample - runStdDev.newMean);
        
                // set up for next iteration
            	runStdDev.oldMean = runStdDev.newMean; 
            	runStdDev.oldStd = runStdDev.newStd;
            }

    }
    
    public static void sample(RunningStdDev runStdDev, long sample) {
    	//important because we must not use isNaN since this stops JIT
    	final double s = sample;
        if (samplerunStdDev.max) {
        	runStdDev.max = s;
        }
        runStdDev.sampleCount++;
        // See Knuth TAOCP vol 2, 3rd edition, page 232
        if (runStdDev.sampleCount == 1)
        {
        	runStdDev.oldMean = runStdDev.newMean = s;
        	runStdDev.oldStd = 0.0;
        } else {
        	double temp = s - runStdDev.oldMean;
			runStdDev.newMean = runStdDev.oldMean + temp/runStdDev.sampleCount;
        	runStdDev.newStd = runStdDev.oldStd + temp*(sample - runStdDev.newMean);
    
            // set up for next iteration
        	runStdDev.oldMean = runStdDev.newMean; 
        	runStdDev.oldStd = runStdDev.newStd;
        }

}
    

    public String toString() {
        try {
            return appendTo(this, new StringBuilder()).toString();
        } catch (IOException e) {
            throw new RuntimeException(e);
        }
        
    }
    
    public static A appendTo(RunningStdDev runStdDev, A target) throws IOException {
       
        Appendables.appendValue(target.append("count:"), sampleCount(runStdDev));
        
        target.append(" mean:").append(Float.toString((float)mean(runStdDev))); //TODO: someday Appendables should also support garbage free float to text
        target.append(" std:").append(Float.toString((float)stdDeviation(runStdDev)));
        target.append(" max:").append(Float.toString((float)maxSample(runStdDev)));
        target.append(" min:").append(Float.toString((float)minSample(runStdDev)));      
        return target;
    }
    
    public static double mean(RunningStdDev runStdDev) {
        return runStdDev.newMean;
    }
    
    public static double maxSample(RunningStdDev runStdDev) {
        return runStdDev.max;
    }
    
    public static double minSample(RunningStdDev runStdDev) {
        return runStdDev.min;
    }
    
    public static double variance(RunningStdDev runStdDev) {
        return ( (runStdDev.sampleCount > 1) ? runStdDev.newStd/(runStdDev.sampleCount - 1) : 0.0 );
    }
    
    public static double stdDeviation(RunningStdDev runStdDev) {
        return Math.sqrt(variance(runStdDev));
    }

    public static long sampleCount(RunningStdDev runStdDev) {
        return runStdDev.sampleCount;
    }
    
    
    
    public static double probabilityDensity(RunningStdDev runStdDev, double x) {
        //inputs
        double variance = variance(runStdDev);
        double stdDev = Math.sqrt(variance);
        double mean = mean(runStdDev);
        
        return probabilityDensityY(x, variance, stdDev, mean);
        
    }

 
    
    
}