training options in a MLP created with deep learning toolbox

Question

Alberto Tellaeche 2022-11-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1858473-training-options-in-a-mlp-created-with-deep-learning-toolbox

回答： Sai Kiran 2022-12-22

Hi all,

I am trying to train a stadard MLP created with deep learning toolbox to classify the digits in the MNIST dataset.

As for my example I do not want to use a CNN, I have flattened the image data, creating with each 28*28 image an input vector of 784 elements.

My code is as follows:

clear; clc;
filenameImagesTrain = 'train-images-idx3-ubyte.gz';
filenameLabelsTrain = 'train-labels-idx1-ubyte.gz';
filenameImagesTest = 't10k-images-idx3-ubyte.gz';
filenameLabelsTest = 't10k-labels-idx1-ubyte.gz';
XTrain = processImagesMNIST(filenameImagesTrain);
YTrain = processLabelsMNIST(filenameLabelsTrain);
XTest = processImagesMNIST(filenameImagesTest);
YTest = processLabelsMNIST(filenameLabelsTest);
sizeX = 28;
sizeY = 28;
XVectorTrain = reshape(XTrain, 28*28, 60000);
multilayer_perceptron = [
    sequenceInputLayer(sizeX*sizeY,"Name","input")
    fullyConnectedLayer(32,"Name","capa 1")
    reluLayer("Name","relu")
    fullyConnectedLayer(10,"Name","capa 2")
    softmaxLayer("Name","softmax")
    classificationLayer("Name","classoutput")];
plot(layerGraph(multilayer_perceptron));
options = trainingOptions("sgdm","Plots","training-progress", ...
        "SequenceLength",sizeY*sizeX,...
        "MaxEpochs",40,"MiniBatchSize",1, ...
        "InitialLearnRate", 0.005,"Momentum",0.9, ...
        "ExecutionEnvironment","auto");
        %"MaxEpochs",40,"MiniBatchSize",8, ...
        %"Shuffle","every-epoch", ...
network = trainNetwork(XVectorTrain,categorical(YTrain),multilayer_perceptron,options);

However, I do not get the network to train:

And this is the result until the end.

I know this example can be done, I have seen similar examples in Keras, but I can not make it work in MATLAB.

I would be very grateful if someone could help me with this issue,

Best regards,

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Sai Kiran 2022-12-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1858473-training-options-in-a-mlp-created-with-deep-learning-toolbox#answer_1133982

Hi,

The function trainNetwork returns the trained model. The returned model gets stored in the variable network, and can be used for predicting the results.

Please refer to the following documentation to know how to use the trained model in further steps.

Train deep learning neural network - MATLAB trainNetwork - MathWorks India

I hope it resolves your query.

Regards,

Sai Kiran Ratna