[elm classification] classification of UCI data sets based on nuclear limit learning machine and limit learning machine, with matlab code

1 Content introduction

Extreme learning machine （ｅｘｔｒｅｍｅ ｌｅａｒｎｉｎｇ ｍａｃｈｉｎｅ,ＥＬＭ） As a new machine learning method , It has become a hot research direction . ＥＬＭ Randomly determine the input weights and hidden layer node offsets of the single hidden layer network , The output weight can be obtained by simple calculation , It not only accelerates the learning speed , And it has good generalization ability . Kernel function ＥＬＭ Introduce kernel function into limit learning machine theory , The least square optimization solution can be obtained , It's solved

ＥＬＭ Algorithm random initialization problem , It has good robustness to model learning parameters , With more stable 、 Better generalization performance .

2 Simulation code

clcclear all​DataSetName = 'waveform';                 % With matlab The name of the dataset stored in the format TrainSamplesPartio = 0.5;                % Proportion of training samples of each type   C = 100;                                 % Regularization factor  DataSet  = load(DataSetName, '-mat');    % Load file into program DataSet = getfield(DataSet, DataSetName);% Get the data in the file KindIdx = DataSet(:,1);                  % Get category number , Data number one 1 List as category number KindCount = length(unique(KindIdx));     % Total number of categories SortedDataSet = sortrows(DataSet, 1);    % Arrange data by class label , Rearrange data from small to large according to class label , Facilitate random data extraction later clear DataSet​EachKindCount = zeros(1,KindCount);         % Number of data of each type for i=1:KindCount    EachKindCount(i) = sum(sum( i == SortedDataSet(:,1)));  % The first i Number of class data end​​TrainData = [];TestData = [];% Put all kinds of training data and test data into a matrix TrainTemp = [];TestTemp = []; for i=1:KindCount    if i==1        TempIdx1(1:TrainSamCountEachKind(i),1) = i;     % Mark the training data class label         TempIdx2(1:size(TestDataEachKind{i},1),1) = i;  % Mark the test data class label         TrainData = [TempIdx1 TrainDataEachKind{1,i}];  % The first column is the class label         TestData = [TempIdx2 TestDataEachKind{1,i}];        TempIdx1 = [];                                  % Clear the class label here , Otherwise, wrong class labels may be generated , Because the number of each type of data may be different         TempIdx2 = [];        TrainTemp = TrainData;        TestTemp = TestData;    else        TempIdx1(1:TrainSamCountEachKind(i),1) = i;        TempIdx2(1:size(TestDataEachKind{i},1),1) = i;        TrainData = [TrainTemp;[TempIdx1 TrainDataEachKind{1,i}]];        TestData = [TestTemp;[TempIdx2 TestDataEachKind{1,i}]];        TrainTemp = TrainData;        TestTemp = TestData;        TempIdx1 = [];        TempIdx2 = [];    endendclear TrainTemp TestTemp TrainDataEachKind TestDataEachKind ​% The statistical results [label_index_actual, label_index_expected, TrainingAccuracy, TestingAccuracy] = elm(TrainData, TestData, 1, KindCount, 'sig');% Format and output relevant information , It can be modified according to your own needs fprintf('Run Information:\n')fprintf('+-----------------------------------------------+\n')fprintf('|KindIdx  | EachKindCount |  EachTrainKindCount |\n');fprintf('+-----------------------------------------------+\n')Msgemat='|  %2d     |%9d      |%11d          |\n';for i=1:KindCount    fprintf(Msgemat, i, EachKindCount(i), TrainSamCountEachKind(i));    fprintf('+-----------------------------------------------+\n')endfprintf('+------------------------------------------------------+\n')figureplot(label_index_actual,'rs')hold onplot( label_index_expected,'bo')xlabel(' sample ')legend(' True value ',' Test value ')title([' Test accuracy =',num2str(TestingAccuracy)])​​​

3 Running results

4 reference

[1] Wang Li , LAN Zhi , Yang Rong , etc. . Based on kernel limit learning machine and wavelet packet transform EEG classification method [J]. Beijing Biomedical Engineering , 2018, 37(5):8.

[2] Wang Li et al . " Based on kernel limit learning machine and wavelet packet transform EEG classification method .". 

About bloggers ： Good at intelligent optimization algorithms 、 Neural networks predict 、 signal processing 、 Cellular automata 、 The image processing 、 Path planning 、 UAV and other fields Matlab Simulation , relevant matlab Code problems can be exchanged by private letter .

Some theories cite network literature , If there is infringement, contact the blogger to delete .