Add random_state

[tlikhomanenko]

Hi!

To reproduce of training it is needed set numpy.random.seed at this moment. Why must I do this (what kind of procedure have random state?) Can you add random_state as parameter for the net initialization?

Thanks, Tatiana.

[zueve]

For reproduce training you can save network state, with functions like neurolab.tool.get_np / set_np:

net_state = nl.tool.get_np(net)
net.tarain(...)
nl.tool_np(net, net_state) # net return to init state

Or make copy all network:

net2 = net.copy()

Note: some algorithms must use random generator

[tlikhomanenko]

I want to reproduce in such way:

I have notebook with net training and I want that notebook gives me the same results for different runnings. And one the hack is to set numpy.random.seed, but I don't do this, because another parts can use the global seed. That is why I need to set random state as a parameter while I am creating net.