基于matlab的卷积神经网络CNN图像分类预测

clc clear close all %% Read % x111, 112 the ones ends in 1 is type1 % [a111,a112,a121,a122]=readdata('sub1A.mat'); % [a211,a212,a221,a222]=readdata('sub2A.mat'); % [a311,a312,a321,a322]=readdata('sub3A.mat'); % [a411,a412,a421,a422]=readdata('sub4A.mat'); % [a511,a512,a521,a522]=readdata('sub5A.mat'); % [a611,a612,a621,a622]=readdata('sub6A.mat'); % [a711,a712,a721,a722]=readdata('sub7A.mat'); % [a811,a812,a821,a822]=readdata('sub8A.mat'); % [a911,a912,a921,a922]=readdata('sub9A.mat'); % [a1011,a1012,a1021,a1022]=readdata('sub10A.mat'); % % %% Reshape % % % TargetData = []; % % TargetData = cat(3, a111, a211, a311, a411, a511, a611, a711, a811, a911, a1011, ... % % a121, a221, a321, a421, a521, a621, a721, a821, a921, a1021); % % % % UntargetData = []; % % UntargetData = cat(3, a112, a212, a312, a412, a512, a612, a712, a812, a912, a1012, ... % %q%2027771338 a122, a222, a322, a422, a522, a622, a722, a822, a922, a1022); % % TargetData = cat(3, a111, a211, a311, a411); % UntargetData = cat(3, a112, a212, a312, a412); % % TargetNum = length(TargetData); % UntargetNum = length(UntargetData); % % N1 = randperm(TargetNum); % TargetData = TargetData(:,:,N1(1:TargetNum)); % % N2 = randperm(UntargetNum); % UntargetData = UntargetData(:,:,N2(1:TargetNum)); % % numTr = floor(0.8 * TargetNum); % numTe = floor(0.8 * TargetNum); % % x1Train = TargetData(:,:,1:numTr); % x2Train = UntargetData(:,:,1:numTe); % x1Test = TargetData(:,:,1+numTr:end); % x2Test = UntargetData(:,:,1+numTe:end); % % xTrain = cat(3, x1Train ,x2Train); % xTest = cat(3, x1Test ,x2Test); % % [~,~,y1Tr] = size(x1Train); % [~,~,y2Tr] = size(x2Train); % [~,~,y1Te] = size(x1Test); % [~,~,y2Te] = size(x2Test); % % yTrain = [1*ones(y1Tr,1) ; 2*ones(y2Tr,1) ]; % yTest = [1*ones(y1Te,1) ; 2*ones(y2Te,1) ]; % save maydata.mat %% Prepare input % [~,~,numTrain ]= size(xTrain); % [~,~,numTest ]= size(xTest); % % X_Train = reshape(xTrain,[64 100 1 numTrain]); % X_Test = reshape(xTest,[64 100 1 numTest]); % % Y_Train = categorical(yTrain); % Y_Test = categorical(yTest); % save maydata.mat Y_Train Y_Test X_Train X_Test %% network params load maydata.mat [M,N,Z,K]=size(X_Train); inputSize = [M,N]; opt.n_chans = 64; opt.n_classes = 2; opt.n_start_chans = 1; opt.n_filters_time=opt.n_start_chans; opt.n_filters_spat=opt.n_start_chans; opt.n_chan_factor = 2; opt.input_time_length= 100; opt.n_filters_2 = (opt.n_start_chans * opt.n_chan_factor); opt.n_filters_3 = (opt.n_start_chans * (opt.n_chan_factor ^ 2.0)); opt.n_filters_4 = (opt.n_start_chans * (opt.n_chan_factor ^ 3.0)); opt.filter_length_2=10; opt.filter_length_3=10; opt.filter_length_4=3; opt.final_conv_length=20; opt.filter_time_length=3; opt.pool_time_stride = 2; opt.conv_stride = opt.pool_time_stride; opt.pool_stride = 1; opt.pool_time_length = 1; opt.drop_prob = 0.05; opt.targetLength = 100; layers = constructNetwork2(opt); analyzeNetwork(layers) % analyzeNetwork(layers1); %% start training options = trainingOptions('sgdm', ... 'InitialLearnRate',0.002, ... 'ValidationData',{X_Test,Y_Test},... 'MaxEpochs',100, ... 'MiniBatchSize',64,... 'ValidationFrequency',100, ... 'ExecutionEnvironment','cpu',... 'L2Regularization',0.02,... 'Plots','training-progress',... 'ValidationPatience', 100); net1 = trainNetwork(X_Train,Y_Train,layers,options); %% confusion matrix [accuracy_Train,accuracy_Test] = accuracy_plotconfusionCNN(net1,X_Train,Y_Train,X_Test,Y_Test);