Deep Learning

In the paper, we provide access toall the coat for the simulation. And we also use niper parameter automatic tunin tool. So every single model, i don't mind if its deeplernin baste or sallow learning base, i soptimise in the same way. What we are trying to say is that you don't have to complicate yourself too much. A few layers, that's three neural lyers. With ten hundred, maybe it is enough. Tis the enth of the road. There is no more. We try with thousand of neurons per lyer and deeper architectures, and the results are not better. Why? Because there are not more