I'm not criticizing your intellectual or mathematical integrity, or your
competence in any way. That is all quite first class in every way. It is deeply
appreciated for those very reasons.
What I have questioned and am now questioning in my own mind is whether
neural networks really add value, in the sense of dollars in hand, over all types
of market cycle environments? I think that individual human beings with great
skill and insight do much better. That is meant not as an "anti-machine" comment
either, but just as an observation over many years of market experience. Perhaps
networks can serve as a baseline or "benchmark", just as moving average systems
can do. In that way they could be a part of a portfolio management system, but
not the whole thing.