how to choose the hidden layers for MLP classifier?

Question

i'm working in a stacking method , i had a feature vector with 2816 features(1536 features from the first model and 1280 from the second model ) and i want to give this vector to an MLPclassifier to give me the final prediction .

i'm asking if there is a method that help me to choose the number of hidden layers and the number of neurones for each one?

Answer 1

Try first using a single input/single hidden/single output layer. You can add more hidden layer after if you see poor performance.