So far, I have seen... Genetic Algorithms Back Propogation What other techniques are there for training NN''s? Any keywords I should be googling for? Thanks

well, all those backpropagation mods like bprop+momentum, quickprop ..

and then maybe backpercolation, cascade correlation, SOM, LVQ, Counterpropagation, jordan/elman nets, hopfield nets, boltzmann machines, ART - nets, neocognitron ...

maybe you should tell us what types of neural networks you wanna use, for which purpose etc.

[edited by - as31415rin on September 10, 2003 5:17:54 AM]

Resilient Propagation, or RProp is truly awesome...

quote: Original post by Raeldor
So far, I have seen...

Genetic Algorithms
Back Propogation

What other techniques are there for training NN''s? Any keywords I should be googling for?

I assume that you mean feedforward neural networks, specifically? Any of the usual numeric optimization techniques (hill climbing, conjugate gradients, simplex, etc.) are fair game, plus any of the newer blind optimization techniques (notably genetic algorithms and simulated annealing, their relatives, and variations). Also, a number of neural network-specific training algorithms have been devised, like CHIR and NOVEL.

See the comp.ai.neural-nets (Usenet) FAQ for more information.