Matlab tips (25) competitive neural network and SOM neural network

Preface

MATLAB Learning about image processing is very friendly , You can start from scratch , There are many encapsulated functions that can be called directly for basic image processing , This series of articles is mainly to introduce some of you in MATLAB Some concept functions are commonly used in routine demonstration ！

Competitive neural network is based on unsupervised learning （Unsupervised learning） Method is an important type of neural network , It is often used as a basic form of network , Form other self-organizing networks , Such as self-organizing mapping network （SOM）、 Adaptive resonance theory network 、 Learning vector quantization network .

Self organizing maps (Self-organizing map, SOM) By learning to input data in space , Generate a low dimension 、 Discrete mapping (Map), To some extent, it can also be regarded as a dimension reduction algorithm .SOM It is an unsupervised artificial neural network . Different from the general neural network, it is trained based on the reverse transfer of loss function , It uses competitive learning (competitive learning) Strategy , Rely on the competition between neurons to gradually optimize the network . And use the nearest neighbor function (neighborhood function) To maintain the topology of the input space . Maintain the topology of the input space ： signify Two dimensional mapping includes the relative distance between data points . Adjacent samples in the input space are mapped to adjacent output neurons .

Due to unsupervised learning , This means that there is no need for human intervention in the training stage ( That is, no sample label is required ), That is, without knowing the category , Clustering the data ; You can identify features that are intrinsically related to a problem . Competitive neural network and SOM A simulation example of neural network prediction, analysis and comparison is shared with you ,MATLAB Version is MATLAB2015b.

One . MATLAB Simulation

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% function ： Competitive neural network and SOM neural network 
% Environmental Science ：Win7,Matlab2015b
%Modi: C.S
% Time ：2022-06-27
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%% I.  Clear environment variables 
clear all
clc

tic
%% II.  Training set / Test set generation 
%%
% 1.  Import data 
load water_data.mat

%%
% 2.  Data normalization 
attributes = mapminmax(attributes);

%%
% 3.  Division of training set and test set 

%  Training set ——35 Samples 
P_train = attributes(:,1:35);
T_train = classes(:,1:35);
%  Test set ——4 Samples 
P_test = attributes(:,36:end);
T_test = classes(:,36:end);

%% III.  Creation of competitive neural network 、 Training and simulation testing 
%%
% 1.  Creating networks 
net = newc(minmax(P_train),4,0.01,0.01);

%%
% 2.  Set training parameters 
net.trainParam.epochs = 500;

%%
% 3.  Training network 
net = train(net,P_train);

%%
% 4.  The simulation test 

%  Training set 
t_sim_compet_1 = sim(net,P_train);
T_sim_compet_1 = vec2ind(t_sim_compet_1);
%  Test set 
t_sim_compet_2 = sim(net,P_test);
T_sim_compet_2 = vec2ind(t_sim_compet_2);

%% IV. SOFM Neural networks create 、 Training and simulation testing 
%%
% 1.  Creating networks 
net = newsom(P_train,[4 4]);

%%
% 2.  Set training parameters 
net.trainParam.epochs = 200;

%%
% 3.  Training network 
net = train(net,P_train);

%%
% 4.  The simulation test 

%  Training set 
t_sim_sofm_1 = sim(net,P_train);
T_sim_sofm_1 = vec2ind(t_sim_sofm_1);
%  Test set 
t_sim_sofm_2 = sim(net,P_test);
T_sim_sofm_2 = vec2ind(t_sim_sofm_2);

%% V.  Results contrast 
%%
% 1.  Competitive neural network 
result_compet_1 = [T_train' T_sim_compet_1']
result_compet_2 = [T_test' T_sim_compet_2']

%%
% 2. SOFM neural network 
result_sofm_1 = [T_train' T_sim_sofm_1']
result_sofm_2 = [T_test' T_sim_sofm_2']
toc

Two . Simulation results

result_compet_1 =

     1     1
     1     4
     1     2
     1     1
     1     1
     1     2
     2     4
     2     4
     2     4
     2     4
     2     4
     2     4
     2     4
     2     4
     2     1
     2     1
     2     4
     2     4
     3     3
     3     3
     3     3
     3     3
     3     3
     3     3
     3     3
     3     3
     3     3
     4     2
     4     2
     4     2
     4     2
     4     1
     4     1
     4     1
     4     1


result_compet_2 =

     1     4
     2     4
     3     3
     4     2


result_sofm_1 =

     1     9
     1    11
     1    12
     1     9
     1     4
     1    12
     2    10
     2    10
     2    10
     2    10
     2    10
     2    10
     2    13
     2    10
     2     9
     2     9
     2    14
     2    10
     3     3
     3     5
     3     1
     3     1
     3     1
     3     2
     3     1
     3     2
     3     1
     4    15
     4    12
     4    16
     4    15
     4     7
     4     7
     4     8
     4     8


result_sofm_2 =

     1    11
     2    10
     3     1
     4    15

 Time has passed  10.817151  second .

In turn, click Plots Medium SOM Topology,SOM Neighbor Connections,SOM Neighbor Distances,SOM Input Planes,SOM Sample Hits,SOM Weight Positions The following simulation diagram can be obtained ：

3、 ... and . Summary

Competitive neural network and SOM Neural networks train , Example simulation of predictive analysis , In fact, in my own column 《MATLAB neural network 43 A case study 》 These two kinds of neural networks are also introduced in , Interested students can also move to the column , Link at the end of the article . Learn one every day MATLAB Little knowledge , Let's learn and make progress together ！

MATLAB neural network 43 A case study