Ok.
Do you think that implementing the model with theano could speed up the code?
I don’t know exactly when theano is used in the background or when it is required to use it in my own code. I don’t know if using the theano exp function instead of the numpy function would improve the computation time, or would change anything.
I understood that np.exp() will use theano.tensor.exp() in the background, is that correct? In that case I guess that there is no benefit to use theano explicitely.