How can I adjust a deep neural network input layer to use a matrix as the input layer?

Question

Lynn Marciano 2018-7-5

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/409037-how-can-i-adjust-a-deep-neural-network-input-layer-to-use-a-matrix-as-the-input-layer

回答： Jahetbe 2022-2-9

I have a matrix of 419632 x 1420 doubles. I tried to create a custom deep neural network model but it seems as though the inputLayers that MATLAB provides are the imageInputLayer and sequenceInputLayer. I tried inputting my data as a sequenceInputLayer but when I did that, it wanted to make my model into a recurrent neural network; which is not what I want to do. Is there anyway to have basically a 'matrixInputLayer' or has anyone encountered this issue and could lead me in the right direction to get good insight to maybe create my custom input layer that handles matrices?

Thanks in advance.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Alpha Bravo 2018-7-7

A black and white image is basically a matrix. You could try to use [419632 1420 1] as the input dimension or somehow fool Matlab into thinking your matrix is a black and white image.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Samuel Louis Sudhof 2019-12-18

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/409037-how-can-i-adjust-a-deep-neural-network-input-layer-to-use-a-matrix-as-the-input-layer#answer_406895

I have the same question. Telling matlab that it's somehow image data when it's not seems hacked. Isn't there a cleaner way to insert raw matrix data into a neural network?

Seems to me like the most basic input layer of a neural network is missing, if that's the case. What gives?

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。

Answer 2

Jahetbe 2022-2-9

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/409037-how-can-i-adjust-a-deep-neural-network-input-layer-to-use-a-matrix-as-the-input-layer#answer_892185

nFeatures = 20;

nExamples = 10000;

nOutputs = 1; % this example is for setting up a regression problem

x = rand(nExamples,nFeatures);

t = rand(nExamples, nOutputs);

XNew = reshape(x', [1,1,size(x,2),size(x,1)]);

layers = [ ...

imageInputLayer([1 1 nFeatures]); % this layer needs the first 3 dimensions of input "XNew"

fullyConnectedLayer(10);

fullyConnectedLayer(nOutputs); % this connected layer needs to have an output-size same as the number of responses (columns) in the output data set "t"

regressionLayer];

options = trainingOptions('sgdm');

trainedNet = trainNetwork(XNew, t, layers, options);