A disk scheduling algorithm based on ant colony optimization (2009) [pdf] | Hacker News Reader

A disk scheduling algorithm based on ant colony optimization (2009) [pdf]

sajjadium.github.io

47 pointssajjadium8 comments

For those interested in optimization based on ant colony behaviour, the book "Ant Colony Optimization" by Dorigo and Stutzle is the text to get. It's quite old now (2004) and unfortunately hasn't seen a second edition. It's a bit of a forgotten classic, in my opinion. I don't work professionally in this area but I found it to be highly approachable. The various ACO algorithms are appealing not only for their utility but also for their intuitiveness. I mean, it's easy to visualise ants cruising around, leaving pheromone trails, and so forth.

Ant colony, like other evolutionary algorithmic such as genetic algorithms, is very effective for solving optimization problems!

Does this need a (2009)?

Good catch! Added.

These "animal based" optimization algorithms are a really weird scene, and based on my brief encounters with it I'm quite sceptical of it. A lot of it is based on "models" of animal behavior, but in reality this is more a marketing gimmic and connections to animal behavior is mostly in the imagination of the authors. And for some reason the animal (and other) metaphors seem to be a substitute for rigorous study of the algorithms' properties.

There's a good elaboration of this here: https://web.archive.org/web/20131102075645/http://antor.ua.a...

I would say it's similar with neural networks and real neurons. They are established ways of referring to the algorithms and I doubt the authors of the paper use them in bad faith

Ant-colony optimization is a search algorithm that is inspired by observing the behaviour individual ants as they search for food and the emergent behaviour of the colony. Ants perform random search, unless they either (1) find food - then they way as direct a way back to the nest as they remember, laying down a trail of pheromones as they walk, or (2) they smell a pheromone trail and then follow it towards the source of food. When the source of food is exhausted, the pheromones evaporate, so they can forget what they have learnt. The emergent behaviour is that they can collectively solve problems such as finding the shortest path to food, or solving the NP-complete travelling salesman problem efficiently.

Reference: http://staff.washington.edu/paymana/swarm/stutzle99-eaecs.pd...

These algorithms are not trying to emulate the behavior of animals, the goal is to use novel search strategies and the animal metaphor just provides a quick way to remember/advertise how it works.

Reply on news.ycombinator.com