Henry Gas Solubility Optimization for Feature Selection

Introduction

• This toolbox offers a Henry Gas Solubility Optimization ( HGSO ) method
• The Main file illustrates the example of how HGSO can solve the feature selection problem using benchmark data-set.

Input

• feat : feature vector ( Instances x Features )
• label : label vector ( Instances x 1 )
• N : number of gas
• max_Iter : maximum number of iterations
• num_clus : Number of gas types
• K : Constant
• alpha : Influence of other gas
• beta : Constant
• L1 : Initial parameter
• L2 : Initial parameter
• L3 : Initial parameter

Output

• sFeat : selected features
• Sf : selected feature index
• Nf : number of selected features
• curve : convergence curve

Example

% Benchmark data set 
load ionosphere.mat; 

% Set 20% data as validation set
ho = 0.2; 
% Hold-out method
HO = cvpartition(label,'HoldOut',ho);

% Parameter setting
N        = 10; 
max_Iter = 100; 
num_clus = 2;      % number of gas types / cluster
K        = 1;      % constant
alpha    = 1;      % influence of other gas
beta     = 1;      % constant 
L1       = 5E-3; 
L2       = 100; 
L3       = 1E-2;

% Henry Gas Solubility Optimization
[sFeat,Sf,Nf,curve] = jHGSO(feat,label,N,max_Iter,num_clus,K,alpha,beta,L1,L2,L3,HO);

% Plot convergence curve
plot(1:max_Iter,curve);
xlabel('Number of iterations');
ylabel('Fitness Value'); 
title('HGSO'); grid on;

Requirement

• MATLAB 2014 or above
• Statistics and Machine Learning Toolbox