Difference between revisions of "Scilab OldKiwi" - Rhea

Revision as of 00:02, 12 March 2008

Here you can find relevant information on how to implement Pattern Recognition projects using Scilab.

Brief Introduction to Scilab

Scilab [1] is a open-source Matlab-like tool developed at INRIA. It can be downloaded for several platforms from the link:

http://www.scilab.org/download/

Tutorials describing how to use Scilab can be found here:

• The official documentation[2]

• A hands-on tutorial [3]

• `ANU Scilab Tutorial [4]

• `Une introduction a Scilab [5] , if you want to have some fun reading a Scilab tutorial in French.

Homework #1 related functionality

Random Number Generator: grand is the function used to generate random numbers. In order to generate a multivariate normally distributed sequence of *n* vectors with mean *mu* and covariance *cov*, grand should be called as:

numbers = grand(n, 'mn',mu, cov);

Function declaration: example that computes the multivariate normal probability density:

// function to compute the multivariate normal distribution //note that it asks for the sigma inverse, as well as the Sigma's determinant function [g] = MultivariateNormalDensity(x,mu, sigma_inv, sigma_det) d=length(x); r2 = (x-mu)'*sigma_inv*(x-mu); factor = 1/sqrt(((2*%pi)^d)*sigma_det); g = factor * exp (-(1/2)*r2); endfunction

The file with the code above can be downloaded from the link below:

Homework #1 related functionality

• Random Number Generator: grand is the function used to generate random numbers. In order to generate a multivariate normally distributed sequence of *n* vectors with mean *mu* and covariance *cov*, grand should be called as:

numbers = grand(n, 'mn',mu, cov);

Function declaration: example that computes the multivariate normal probability density:

// function to compute the multivariate normal distribution //note that it asks for the sigma inverse, as well as the Sigma's determinant function [g] = MultivariateNormalDensity(x,mu, sigma_inv, sigma_det) d=length(x); r2 = (x-mu)'*sigma_inv*(x-mu); factor = 1/sqrt(((2*%pi)^d)*sigma_det); g = factor * exp (-(1/2)*r2); endfunction

The file with the code above can be downloaded from the link below:

!`MultivariateNormalDensity.sci`__

__ MultivariateNormalDensity.sci

Homework #1 related functionality

• Random Number Generator: grand is the function used to generate random numbers. In order to generate a multivariate normally distributed sequence of *n* vectors with mean *mu* and covariance *cov*, grand should be called as:

numbers = grand(n, 'mn',mu, cov);

• Function declaration: example that computes the multivariate normal probability density:

// function to compute the multivariate normal distribution //note that it asks for the sigma inverse, as well as the Sigma's determinant function [g] = MultivariateNormalDensity(x,mu, sigma_inv, sigma_det) d=length(x); r2 = (x-mu)'*sigma_inv*(x-mu); factor = 1/sqrt(((2*%pi)^d)*sigma_det); g = factor * exp (-(1/2)*r2); endfunction

The file with the code above can be downloaded from the link below:

!`MultivariateNormalDensity.sci`__

__ MultivariateNormalDensity.sci

Tool Boxes

There are several tool boxes of functions written by people all over the world adding extra functionality to Scilab. Here are some useful links:

• Toolboxes for Scilab and Their Manuals [6]

• Scilab Toolbox especialized on Pattern Recognition -- Presto-Box [7]

• Manual for Presto-Box -- Scilab [8]

• ANN - Neural Networks Tool Box [9]

• SCIsvm [10] , a plugin for the libsvm C++ library.

Scilab Code

All the relevant code for the EE662 course written in Scilab is posted below:

!`MultivariateNormalDensity.sci`__ - Implementation of a function to compute the multivariate normal density

__ MultivariateNormalDensity.sci