(April 22, 2015 at 5:23 am)Alex K Wrote: I think when you'd train an artificial neural network, you would also slowly reduce the factor by which new inputs rescale the weights, once the most important stuff is learned. That might mimic the effect you mention to some extent.
Might, yes... but this is a self-training NN, we're talking about here... would that self-training work the same way?... Maybe if it was "genetically" encoded...
