I found a small code snippet on machine precision for float and double datatypes in C….

ani@ani-desktop:~/Desktop/executedfiles$ javac retrievalGabor.java
retrievalGabor.java:54: incompatible types
found : double[]
required: float[]
query= new double[] {0.1544465f ,0.54298925f ,0.73340774f ,0.017640665f ,0.085013896f ,0.095653236f ,0.0 ,0.4047619f ,0.36394557f ,0.21768707f ,0.09863946f ,0.088435374f ,0.08163265f ,0.06462585f ,0.05782313f ,0.27210885f ,0.26870748f ,0.10204082f ,0.0 ,0.010204081f ,0.0034013605f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.023809524f ,0.010204081f ,0.0 ,0.0 ,0.17687075f ,0.32312927f ,0.0 ,0.0 ,0.09863946f ,0.6632653f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.013605442f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.08163265f ,0.0 ,0.0 ,0.0 ,0.49659863f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.010204081f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0};
^
1 error

code here:

//The alignment of the features is assumed to be AVG, VAR, CSD, GABOR

import java.io.*;
import javax.imageio.*;
import java.awt.*;
import java.awt.image.*;

class retrievalGabor
{
public static void main(String args[])
{
if(args.length!=6 && args.length!=7)
{
System.out.println(“The parameters are {Query Image} {Query Gabor file} {Features File} {Gabor Text File} [blocks] {Relevant Features} {EPS directory}”);
System.exit(0);
}

int BLOCKS=1, FEATURES=15, SCALE=3, ORIENTATION=4;
BufferedImage bi, bi1;
GraphicsEnvironment ge= GraphicsEnvironment.getLocalGraphicsEnvironment();
GraphicsDevice gd=ge.getDefaultScreenDevice();
GraphicsConfiguration gc=gd.getDefaultConfiguration();
Graphics2D g2D=null;
ColorModel cm=null;
DataInputStream dis, dis1;
BufferedReader br, br1;

float features[][], diff[], query[], tmp=0;
String names[], tmps=null, tokens[];
int images=0, wt[], bitcount=0;
char ch;

try{
bi=ImageIO.read(new File(args[0]));
bi1=gc.createCompatibleImage(bi.getWidth(null), bi.getHeight(null));
g2D=bi1.createGraphics();
g2D.drawImage(bi, null, 0, 0);
bi=bi1;

dis=new DataInputStream(new FileInputStream(args[2]));//opening the features file
br=new BufferedReader(new FileReader(args[3]));//opening the texture features file

if(args.length==7)
{
BLOCKS=Integer.parseInt(args[4]);
dis1=new DataInputStream(new FileInputStream(args[5]));//opening the relevant features file
}
else
dis1=new DataInputStream(new FileInputStream(args[4]));//opening the relevant features file

//query=myFinalToolkit.computeColorSampleImage(bi, BLOCKS, SCALE*ORIENTATION);
query= new double[] {0.1544465f ,0.54298925f ,0.73340774f ,0.017640665f ,0.085013896f ,0.095653236f ,0.0 ,0.4047619f ,0.36394557f ,0.21768707f ,0.09863946f ,0.088435374f ,0.08163265f ,0.06462585f ,0.05782313f ,0.27210885f ,0.26870748f ,0.10204082f ,0.0 ,0.010204081f ,0.0034013605f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.023809524f ,0.010204081f ,0.0 ,0.0 ,0.17687075f ,0.32312927f ,0.0 ,0.0 ,0.09863946f ,0.6632653f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.013605442f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.08163265f ,0.0 ,0.0 ,0.0 ,0.49659863f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.010204081f ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0 ,0.0};

br1=new BufferedReader(new FileReader(args[1]));//opening the Query texture file
tokens=(br1.readLine()).split(” “);
for(int i=3;i<2*BLOCKS*SCALE*ORIENTATION;i++)
query[BLOCKS*6+120+i-3]=Float.parseFloat(tokens[i]);
br1.close();

//reading the features file
images=dis.readInt();
features=new float[images][BLOCKS*(6+SCALE*ORIENTATION*2)+120];
names=new String[images];

for(int i=0;i<images;i++)
{
names[i]=new String();
while((ch=dis.readChar())!=’\n’)
names[i]+=ch;

for(int j=0; j<BLOCKS*6+120;j++)
features[i][j]=dis.readFloat();
dis.readChar();

tokens=(br.readLine()).split(” “);
for(int j=3;j<BLOCKS*2*SCALE*ORIENTATION;j++)
features[i][BLOCKS*6+120+j-3]=Float.parseFloat(tokens[j]);
}
dis.close();
br.close();

//reading the feature relevancy
wt=new int[BLOCKS*(FEATURES-1)+1];
for(int i=0;i<BLOCKS*(FEATURES-1)+1;i++)
{
wt[i]=dis1.readInt();
System.out.print(wt[i]+” “);
}
System.out.println();
dis1.close();

diff=new float[images];
for(int i=0;i<images;i++)
{
//average and variance in color
for(int j=0;j<BLOCKS;j++)
{
for(int k=0;k<2;k++)
{
tmp=0;
if(wt[j*(FEATURES-1)+k]==1)
{
for(int l=0;l<3;l++)
tmp+=Math.abs(query[j*6+k*3+l]-features[i][j*6+k*3+l]);

diff[i]+=tmp/3;
}
}
}

//CSD
tmp=0;
if(wt[BLOCKS*(FEATURES-1)]==1)
{
for(int k=0;k<120;k++)
tmp+=Math.abs(query[BLOCKS*6+k]-features[i][BLOCKS*6+k]);

diff[i]+=tmp/120;
}

//texture features
for(int j=0;j<BLOCKS;j++)
for(int k=0;k<SCALE*ORIENTATION;k++)
{
tmp=0;
if(wt[j*(FEATURES-1)+2+k]==1)
for(int l=0;l<2;l++)
{
tmp+=Math.abs(query[BLOCKS*6+120+j*SCALE*ORIENTATION+2*k+l]-features[i][BLOCKS*6+120+j*SCALE*ORIENTATION+2*k+l]);
}
diff[i]+=tmp/2;
}

}

//sort the entries in the sorted order
for(int i=0;i<images-1;i++)
for(int j=0;j<images-i-1;j++)
{
if(diff[j+1]<diff[j])
{
tmp=diff[j];
diff[j]=diff[j+1];
diff[j+1]=tmp;

tmps=names[j];
names[j]=names[j+1];
names[j+1]=tmps;

tmps=null;
}
}

/* //sort the entries in the sorted order by name
for(int i=0;i<29;i++)
for(int j=0;j0)
{
tmp=diff[j];
diff[j]=diff[j+1];
diff[j+1]=tmp;

tmps=names[j];
names[j]=names[j+1];
names[j+1]=tmps;

tmps=null;
}
}
*/

Runtime rt=Runtime.getRuntime();
String arr[]=new String[3], dir[]=new String[2];

dir[0]=new String(“/bin/mkdir”);
if(args.length==6)
dir[1]=new String(“/tmp/”+args[5]+args[0].substring(args[0].lastIndexOf(‘/’)+1, args[0].lastIndexOf(‘.’)));
else
dir[1]=new String(“/tmp/”+args[6]+args[0].substring(args[0].lastIndexOf(‘/’)+1, args[0].lastIndexOf(‘.’)));

rt.exec(dir);

arr[0]=new String(“/bin/cp”);
for(int i=0;i<30;i++)
{
System.out.println(names[i]+” “+diff[i]);
arr[1]=new String(args[0].substring(0,args[0].lastIndexOf(“/”)+1)+names[i]);

if(i<10)
arr[2]=new String(dir[1]+”/0″+i+”.jpg”);
else
arr[2]=new String(dir[1]+”/”+i+”.jpg”);
rt.exec(arr);
}

}
catch(Exception e)
{
e.printStackTrace();
System.exit(0);
}
}
}

n=10000000;
h=1/(n-1);
a = zeros(n,1);
b=[0:n-1]‘*h;
for i=2:n
a(i) = a(i-1)+h;
end

error = a-b;

plot([0:n-1]‘, error);
title ‘Error a_i-b_i, a_i=a_{i-1}+h, b_i=i*h’
xlabel ‘i’
ylabel ‘error’

This code has been written so that the final value for a and b is 1. Therefore, the error in this result will give you an idea of how many digits of error you can expect for different values of n.

Doubles have about 16 digits of precision. After 10 million iterations, you’ve lost 6 digits of precision. In floating point arithmetic, you will have lost almost all your digits of precision.

Here’s the plot:

double inc = 1.e-5;
for (int iter=0; iter < 900; iter++) {
double dnumber = inc*iter;
}