demobdk.m

有监督自组织映射-偏最小二乘算法(A supervised self-organising map–partial least squares algorithm),可以用语多变量数据的回归分析

function DemoBDK;
%
clear
close all
% make a copy of the parameter file
disp('copying demo-parameters for the Wine data set to SetParamsModel.m');
!copy SPMBDK.m SetParamsModel.m
% build a BDK model on basis of a Wine class data set
%
Ninit=1;         % just one random initialisation of the network weights
%                  note that the predifined division in training and test
%                  set is used.
PercTrain=0;     % here ignored, because training and test set are already defined
[PerfTrain,PerfTest,BDK,MLKP] = MakeModel(Ninit,PercTrain);
disp(' ')
disp('!!! hit a key to continue !!!')
pause
%
% PerfTrain contains the percentage correct (for classification) or the total
% RMSEP value for the training set
%
% PerfTest contains the percentage correct (for classification) or the total
% RMSEP value for the test set
%
% BestBDK is a Matlab struct containing all the essentials for the model
%
% MLKP contains the complete parameter block for the BDK model
%
%
% start a quick analysis of the content of BDK
AnalyseBDK(BDK,MLKP);
%
% figure(1) displays the weights of the input map
%
% figure(2) shows in colour the class membership (embedded in the output map)
%
% figure(3) depicts a correlation image between input and output map