how can I replace the softmax layer with another classifier as svm in convolution network

Question

Suheer Ali 2018-4-16

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/395324-how-can-i-replace-the-softmax-layer-with-another-classifier-as-svm-in-convolution-network

评论： amel yasser 2022-5-24

I made deep learning application that using softmax

layers = [ imageInputLayer(varSize); conv1; reluLayer;

    convolution2dLayer(5,32,'Padding',2,'BiasLearnRateFactor',2);
    reluLayer()
    maxPooling2dLayer(4,'Stride',2);
    convolution2dLayer(5,32,'Padding',2,'BiasLearnRateFactor',2);
    reluLayer()
    maxPooling2dLayer(2,'Stride',2);
    convolution2dLayer(5,64,'Padding',2,'BiasLearnRateFactor',2);
    reluLayer();
    maxPooling2dLayer(4,'Stride',2)
    fc1;
    reluLayer();
    fc2;
    reluLayer();
    %returns a softmax layer for classification problems. The softmax layer uses the softmax activation function.
    softmaxLayer()
    classificationLayer()];

I want to use SVM and random forest classifiers instead of softmax. and use a deep learning for feature extraction. I hope I can get a link for a tutorial.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

mona benhari 2021-2-28

I have the same problemI have you got the answer?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Johannes Bergstrom 2018-4-17

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/395324-how-can-i-replace-the-softmax-layer-with-another-classifier-as-svm-in-convolution-network#answer_315540

Here is an example: https://www.mathworks.com/help/nnet/examples/feature-extraction-using-alexnet.html

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Suheer Ali 2018-4-17

Thanks for your answer but I don't want to use pre-trained models. I want to design mine and use it as a feature extraction.

请先登录，再进行评论。

Answer 2

Nagwa megahed 2022-4-21

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/395324-how-can-i-replace-the-softmax-layer-with-another-classifier-as-svm-in-convolution-network#answer_947700

the only possible solution is to save the extracted features by the deep model , then use this features as an input to the SVM or any other wanted classifier.