ndhistc

Multi dimensional histogram

Description

function qHistogram = ndHistc (mData, vEdge1, vEdge2, ... )
 
  * Input Arguments:
 
       + mData: nRecord by nDim 2-dimensional array of doubles
       + vEdge1, vEdge2, ... : nDim vectors of histogram edges
 
  * Return value:
 
       + qHistogram: nDim-dimensional data cube
            containing number of points in each cell
            defined by histogram edges. For instance,
            qHistogram(1,1,1,...) means number of data points
            satisfying
                vEdge1(1) <= mData(:,1) < vEdge1(2) & ...
                vEdge2(1) <= mData(:,2) < vEdge2(2) & ...
                vEdge3(1) <= mData(:,3) < vEdge3(1) & ...
                ...
  * Example
        mRand = rand(1e6,5);
        ve1 = linspace(0,1,5);
        ve2 = linspace(0,1,6);
        ve3 = linspace(0,1,7);
        ve4 = linspace(0,1,8);
        ve5 = linspace(0,1,9);
        qHist = ndhistc(mRand, ve1, ve2, ve3, ve4, ve5);
       
  * Comparison with ndhist.m (compiled using mcc -x ndhist)
       + 1e6 by 2 data -> 5 by 6
           ndhist.m 79.49 sec
           ndhistc.c 0.4610 sec
 
       + 1e6 by 5 data -> 5 by 6 by 7 by 8 by 9
           ndhist.m 199.32 sec
           ndhistc.c 2.4430 sec
 
       ==> More efficient if More data points (==rows) & Less dimensions (==columns)

• MATLAB release: MATLAB 6.1 (R12.1)
• Other requirements: Zip file contents: ndHistc .m .c .dll <- C Mex version ndHist .m .dll <- m file version. Slower. TestndHist.m <- example

[Nelson Martins]

This program is very useful in multi-variate time-series analysis. The MEX implementation is a substantial benefit for the users.
The program is very intuitive, and the MEX works well on 64 bit platform(s).

[Stefan Gachter]

Nice and fast function!

However, there is a small bug on line 221 that can cause problems. The variable vnDimIn has to be removed from the if statement, otherwise "Memory allocation error (1)" can occur for no reasons.

[Kangwon Lee]

Code hosted @ http://dev.naver.com/projects/ndhist/ or svn checkout --username anonsvn https://dev.naver.com/svn/ndhist (pwd anonsvn)