宽度学习 Broad Learning System MATLAB 代码1：MNIST实践

%%%%%%%%%%%%%%%%%%%%%%%%This is the demo for the bls models including the
%%%%%%%%%%%%%%%%%%%%%%%%proposed incremental learning algorithms. 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%load the dataset MNIST dataset%%%%%%%%%%%%%%%%%%%%
clear; 
warning off all;
format compact;
load mnist;

%%%%%%%%%%%%%%%the samples from the data are normalized and the lable data
%%%%%%%%%%%%%%%train_y and test_y are reset as N*C matrices%%%%%%%%%%%%%%
train_x = double（train_x/255）;
train_y = double（train_y）;
% test_x = double（train_x/255）;
% test_y = double（train_y）;
test_x = double（test_x/255）;
test_y = double（test_y）;
train_y=（train_y-1）*2+1;
test_y=（test_y-1）*2+1;
assert（isfloat（train_x） ‘train_x must be a float‘）;
assert（all（train_x（:）>=0） && all（train_x（:）<=1） ‘all data in train_x must be in [0:1]‘）;
assert（isfloat（test_x） ‘test_x must be a float‘）;
assert（all（test_x（:）>=0） && all（test_x（:）<=1） ‘all data in test_x must be in [0:1]‘）;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp（‘Press any key to run the one shot BLS demo‘）;
pause

%%%%%%%%%%%%%%%%%%%%This is the model of broad learning sytem with%%%%%%
%%%%%%%%%%%%%%%%%%%%one shot structrue%%%%%%%%%%%%%%%%%%%%%%%%
C = 2^-30; s = .8;%the l2 regularization parameter and the shrinkage scale of the enhancement nodes
N11=10;%feature nodes  per window
N2=10;% number of windows of feature nodes
N33=11000;% number of enhancement nodes
epochs=10;% number of epochs 
train_err=zeros（1epochs）;test_err=zeros（1epochs）;
train_time=zeros（1epochs）;test_time=zeros（1epochs）;
% rand（‘state‘67797325）   % 12000 %%%%% The random seed recommended by the
% reference HELM [10].
N1=N11; N3=N33;  
for j=1:epochs    
    [TrainingAccuracyTestingAccuracyTraining_timeTesting_time] = bls_train（train_xtrain_ytest_xtest_ysCN1N2N3）;       
    train_err（j）=TrainingAccuracy * 100;
    test_err（j）=TestingAccuracy * 100;
    train_time（j）=Training_time;
    test_time（j）=Testing_time;
end
save （ [‘mnist_result_oneshot_‘ num2str（N3）] ‘train_err‘ ‘test_err‘ ‘train_time‘ ‘test_time‘）;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp（‘Press any key to run the one shot BLS demo with BP algorithm‘）;
pause

%%%%%%%%%%%%%%%%%%%%This is the model of broad learning system for one%%%%%%
%%%%%%%%%%%%%%%%%%%%shot structrue with fine tuning under BP algorithm%%%%%%%%%%%%%%%%%%%%%%%%
C = 2^-30; s = .8;%the l2 regularization parameter and the shrinkage scale of the enhancement nodes
N11=10;%feature nodes  per window
N2=10;% number of windows of feature nodes
N33=5000;% number of enhancement nodes
epochs=1;% number of epochs 
train_err=zeros（1epochs）;test_err=zeros（1epochs）;
train_time=zeros（1epochs）;test_time=zeros（1epochs）;
% rand（‘state‘67797325）   % 12000 %%%%% The random seed recommended by the
% reference HELM [10].
N1=N11; N3=N33;  
for j=1:epochs   
    [TrainingAcc

 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----

     文件       9687  2017-11-27 22:51  Demo_Broadlearning_MNISTBLS_demo_MNIST.m

     文件       9752  2017-11-27 22:52  Demo_Broadlearning_MNISTBLS_demo_MNIST_for_lower_memory.m

     文件       3028  2017-07-16 19:35  Demo_Broadlearning_MNISTls_train.m

     文件       7776  2017-07-19 23:17  Demo_Broadlearning_MNISTls_train_bp.m

     文件       5834  2017-07-16 19:43  Demo_Broadlearning_MNISTls_train_enhance.m

     文件       6745  2017-07-16 20:28  Demo_Broadlearning_MNISTls_train_enhancefeature.m

     文件       5639  2017-07-16 20:31  Demo_Broadlearning_MNISTls_train_input.m

     文件       6398  2017-07-16 20:32  Demo_Broadlearning_MNISTls_train_inputenhance.m

     文件   14735220  2017-07-11 20:58  Demo_Broadlearning_MNISTmnist.mat

     文件        438  2017-07-19 23:02  Demo_Broadlearning_MNISTpre_zca.m

     文件         72  2017-07-11 20:58  Demo_Broadlearning_MNIST
esult.m

     文件        495  2017-07-11 20:58  Demo_Broadlearning_MNISTsparse_bls.m

     目录          0  2017-11-27 22:52  Demo_Broadlearning_MNIST

----------- ---------  ---------- -----  ----

             14791084                    13