Can you have the same amount of inputs than outputs?

I would like to create a 1D CNN. My data is basically composed of thousands of 1D array of 100 points. I would like to extract 1 particular point out of these 100. My plan was to build a neural network where I would have 100 outputs with the probabilities of each points.
Thank you

Hi @Raphael_Henri :wave: Have you tried building a simple Keras Sequential model and running tf.keras.Model.summary to check the shapes? E.g. the Create the convolutional base step section in Evrişimsel Sinir Ağı (CNN)  |  TensorFlow Core but with a softmax layer for probabilities.

Also, check out this example:

1 Like

Hi @8bitmp3,
Thank you for your reply. I will try to do what you are proposing but my main concern is that I would need the final dense layer to have 100 outputs and get the probability for each output. Then I could take the output with the max probability. My concern about this is the precision. I’m not sure it is a good practice to have a ratio of 1:1 for inputs:outputs, I just don’t know and there is not a lot of documentation on this topic. Ideally, my final layer would have one output with the index between 0 and 100. Maybe the two approaches are similar (if the second approach even exists)… What are you thoughts about this?
Thanks for your help again