✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，matlab项目合作可私信。
🍎个人主页：Matlab科研工作室
🍊个人信条：格物致知。
更多Matlab仿真内容点击👇
⛄ 内容介绍
时间卷积网络(TemporalConvolutionalNetwork,TCN)是一种特殊的卷积神经网络(CNN).
① TCN结构中的卷积具有因果关系,如图3(a)所示,数据的传递是单向的,即每层的信息只依赖于之前层的信息。
② 传统卷积神经网络对时间的建模长度受限于卷积核的大小,需要堆叠较多层来抓取更长的依赖关系。
为解决该问题,如图3(b)所示,TCN采用膨胀卷积,即在因果卷积的基础上间隔采样输入。膨胀卷积使得输入信息量随层数的增加呈指数型增长,卷积网络就可以凭借较少的层来获得足够大的感受野。
⛄ 部分代码
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%% This is a function to define the layers and parameters of the CNN %%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [layers, options] = raceNetFun(inputSize, tbl, augmentedTestSet, ep)
layers = [
imageInputLayer([inputSize 3])
convolution2dLayer(11,4,'Padding','same')
batchNormalizationLayer
reluLayer
maxPooling2dLayer(2,'Stride',2)
convolution2dLayer(7,8,'Padding','same')
batchNormalizationLayer
reluLayer
maxPooling2dLayer(2,'Stride',2)
convolution2dLayer(5,16,'Padding','same')
batchNormalizationLayer
reluLayer
maxPooling2dLayer(2,'Stride',2)
convolution2dLayer(3,64,'Padding','same')
batchNormalizationLayer
reluLayer
fullyConnectedLayer(height(tbl))
softmaxLayer
classificationLayer];
options = trainingOptions('adam', ...
'InitialLearnRate',1e-2, ...
'LearnRateSchedule','piecewise', ...
'LearnRateDropFactor',0.5, ...
'LearnRateDropPeriod',5, ...
'MaxEpochs',ep, ...
'Shuffle','every-epoch', ...
'ValidationData',augmentedTestSet, ...
'ValidationFrequency',10, ...
'MiniBatchSize', 64, ...
'Verbose',false, ...
'Plots','training-progress');
end
⛄ 运行结果
⛄ 参考文献
[1]冯达智. 基于时间卷积网络的飞控时序数据预测技术研究[D]. 电子科技大学.
