Quantcast
Channel: Comments on: Neural Darwinism – An Idea Reborn?
Viewing all articles
Browse latest Browse all 2

By: Taliesin Beynon

0
0

I don’t share your enthusiasm, though I’m too stingy with my dollars to fund the same research twice by paying for an article that has already been funded by the taxpayer.

So now that I’m comfortable in my armchair, allow me to ask: how can neural circuits calculate fitness in a local way without solving the problems they are meant to be evolving solutions to in the first place? Usually the rest of the world handles that part.

In fact, that was a similar problem with the famous “backpropogation” algorithm for training neural networks — while possesing of many elegant mathematical properties that allow computers to efficiently train neural nets to solve practical problems, it seems clear that the brain cannot use back propogation in practice because it is not a local algorithm, and so its importance to theoretical neuroscience is minimal.

For an example of theoretical neuroscience having IMHO some good contact with reality, check out this fascinating talk on the interdisciplinary mongrel known as “neural field theory” that elegantly explains cave paintings and acid trips using ideas from particle physics: http://bit.ly/aadyHF


Viewing all articles
Browse latest Browse all 2

Latest Images

Trending Articles





Latest Images