Neural network architecture for multi-output classification

Question

I'm trying to build a neural-network based classifier that takes as input a (32, m) matrix, where 32 is the number of features and m is the batch size, and outputs a (16, m) matrix, where each column is an output vector (16, 1). Each element of a single output vector may have 7 different values, where each one corresponds to a class. How can I build this model in Python using Keras?

Thanks in advance!

Image of what I want to implement

Answer 1

Batch size is a hyperparameter that define number of sample to work through before updating internal model parameters. To design the architecture of your network you do not need information about batch size.

I do not know what kind of the input/output data you are dealing with, but based only on your question, one implementation is with the input layer having 32 neurons and the output layer having 16 (number of outputs) * 7 (number of possible classes) neurons.