Hi,

As author and maintainer of AI-NNFlex, I'd like to invite comment, suggestion & feedback in this forum.

I'm currently looking at ways to speed the code up by implementing C or PDL based modules without breaking the flexibility of it.
I'm determined to maintain AI-NNFlex as a resource for implementing neural networks in pure perl, AI-NNFlex will /always/ have full functionality implemented in perl - Inline/xs/pdl functionality will be implemented to provide fast use, but will be faster optional versions of pure perl code.

As per my comments on perlmonks, I'm happy to invite feedback based purely on a review of perldoc, but if posting on that basis please state that this is the case.

Great! Sounds like an opportunity to improve the module and improve my perl. I'd be grateful for any code optimisation you've made to this. I've done a lot of optimisation since the early versions (chiefly getting rid of eval's and unnecessary function calls, and implementing various tricks (courtesy of Scott Fahlman) for speeding up learning) and recent versions are dramatically faster, but you may well have spotted things I've missed.

After something of a hiatus, while I finished off a contract I was working on and finished writing my MSc dissertation, I've got some time to work on this now. Expect a genetic learning module, perhaps a quickprop module, and maybe even a GUI over the next few weeks :-)

The email address listed in the docs should work, but occasionally people have DNS resolution problems with it (sadly, the spammers rarely have a problem). Alternatively you can reach me on colbourn1ATuk2.net, but make sure your subject line references NNFlex, because I spam filter that address rather aggressively. I can also be reached as g0n on perlmonks, and I'm there for at least a couple of hours a day.

thanks,

Charles