Re:The quest to evolve neural networks through evolutionary ...
by catamorphismic on Monday October 30, 2017 @05:38AM
(#15583884)
> Not necessarily. Everything points to the fact that the human brain (at least, I don't know much about ants) does not work in any way similar to neural networks, and as such there's no guarantee that you can represent its behaviour by the current algorithms.
Yes, the brain is most likely not simply modelling one giant ANN, but it's much more plausible that ANN-like components have a role in it.
Re:The quest to evolve neural networks through evolutionary ...
by canjobear on Monday October 30, 2017 @10:14AM
(#15585119)
If there is a space of optimal policies for the ant, then both the ANN and real neural systems will be attracted there. There are certainly a lot of local minima, but it would be surprising to me if ANNs and real neural systems could only find totally disjoint local minima with no similarities.
Re:The quest to evolve neural networks through evolutionary ...
by zardo on Monday October 30, 2017 @01:24PM
(#15586953)
>and as such there's no guarantee that you can represent its behaviour by the current algorithms.
Doesn't the universal function approximation theorem provide just such a guarantee? It doesn't guarantee any algorithm will converge on that representation, but the capability to represent it is there.
Not necessarily. Everything points to the fact that the humain brain (at least, I don't know much about ants) does not work in any way similar to neural networks, and as such there's no guarantee that you can represent its behaviour by the current algorithms.
For example, real neurons have no supervision signal. Memory is also a big issue (see the work being done by DeepMind and FAIR on differentiable neural computers, and memory networks) and Reinforcement Learning still struggles with long-term planing, switching strategies, etc.