How can I pass raw data as input to a convolutional neural network?

Hi everyone,

I am trying to pass as input a numeric vector of data to a convolutional neural network. I know that it's possible to pass a numeric input to a convolutional network in Python but, do you know how can I do in Matlab?

Answer 1

[Sarah Mohamed]

If I understand correctly, you are interested in specifying an input layer for a CNN that accepts a numeric vector. This is indeed possible using the “imageInputLayer” function:

https://www.mathworks.com/help/nnet/ref/imageinputlayer.html

n = 1000         % n = length of the input signal
inputLayer = imageInputLayer([1 n])  % Input is a row vector of length n
          % Valid type = single or double

You might find the following Answers post relevant to you – this user’s question involves constructing a CNN with a 1D input signal:

https://www.mathworks.com/matlabcentral/answers/331164-convolutional-1d-net

[Diego Alonso]

Thanks for your answer. I have read the post you passed but I have some doubts. I am trying to follow this thesis (chapter 4) http://lemt.ufrj.br/pdf/pedro.pdf.

In summary, I have to disaggregate (into appliances) an aggregated power signal. I have both, aggregated and disaggregated (consumption of each appliance) power signal so I have to pass as input the whole consumption (aggregate) and as target the consumption of an appliance.

How can I do this?