What is an Ant colony optimization algorithm?

What is Ant Colony Optimization and how does it work in layman's terms?

• Answer:

  The ideas is based on an ant colony(hence the name). At first, ants have no idea where food is, and so they run around in every direction. Eventually, some ant stumbles upon some amount of food, and it brings it back home. It leaves a pheremone trail in the process indicating to other ants that there is food there. This pheremone trail makes other ants more likely to go down that route, where they will also find food, drag it back to the colony, and reinforce the trail, causing even more ants to go down that route... For machine learning, you do something similar. You create many agents with different approaches, and they will communicate their success to the other agents, making those agents more likely to try a similar approach. It essentially lets you sample a wide swath of the possible answer space to feel it out, then focus in on the most promising areas you find.