Talk:Ant colony optimization
From Wikipedia, the free encyclopedia
Contents |
[edit] related methods
I suggest to move the related methods section to the "combinatorial optimization" article, also including a short description for ACO.
[edit] algorithm
This article would benefit from a description of the algorithm in Dorigo's paper, in addition to the high-level overview. I can't find information on the implentation of this algorithm anywhere. Twelvethirteen 08:17, 28 May 2006 (UTC)
[edit] Neural networks
Is there any non-trivial difference between ACO and neural networks?
- They're fundamentally different creatures altogether.
- ACO is population-based, ANNs are not.
- ACO is specifically for graph-based problems (or problems which can be converted into graph/decision problems); ANN is more akin to a mathematical function.
- ANN typically trains via BackProp and gradient descent; via computing errors and filtering those errors back. ACO essentially balances numerous greedy aspects with eachother, premised upon a series of continuous values that have no direct connection to 'errors' at all.
- ACO could easily be used (and has been used) to train a neural network, but I can't conceive of any way that an ANN could ever 'train' an ACO or ACS. 139.57.100.104 (talk) 02:44, 3 May 2008 (UTC)
[edit] Update References
Should the references on this article be updated to follow general Wikipedia style guidelines? --Jeffrey Sharkey 18:59, 1 September 2006 (UTC)
[edit] external links
add http://whoyouknow.co.uk/ants/ - currently in progress computer science bsc project investigating ant behaviour and its uses in computing, implemented using java.