I would like to know if someone has an experience with NeuroLab using in there outscript an unstable indicator like ZigZag.

Do you think that we can create a realistic NNIndicator like that ?

Zigzag isn't really "unstable"; the problem with using ZZ in trading system rules is that it peeks, i.e., you need forward looking data to know where the swing highs and lows occurred.

Consequently, you can't use ZZ in the Input Script but it might be useful for training a trend-reversal system in the Output Script. Not sure exactly how you'd do it, but it's an interesting idea to think about.

You'd probably want a momentum indicator like MACD or RSI in the input, and you can use it in the trading script to detect divergences, which moving improve success for counter-trend trades.

Thank you very much. Just another question. Is there a plan to introduce different type of Neural network in NeuroLab like Fuzzy Logic, Dynamic learning vector quantization, Kohonen... ?

Not that I know of.

It’s really shame because trading industries use more and more neural networks in their trading strategies and WLD is not up to date on this part. So, it could be a good thing that WLD can evolve on this way too.

Moreover, there are free library of neural networks that we can found like http://neurondotnet.freehostia.com/index.html or http://www.codeproject.com/KB/recipes/aforge_neuro.aspx to help developers to boost the integration in NeuroLab.

QUOTE:
more and more

Like it or not, it's not possible to pack every known trick in the book into WL/addins for different reasons e.g. limited manpower and time and/or not enough bang for the buck.

QUOTE:
free library

Pre-release version of Neuro-Lab for WL6 was initially based on the "free" NeuronDotNet library (your first link) but a GPLv3 licensing catch could basically force MS123 to disclose the NL source code. Neuro-Lab was rewritten and uses an in-house, custom built NN engine. That was a good lesson to learn: GPL v3 sucks.