Transition behavior of one Artificial Ant Colony in presence of a sudden change in his artificial digital image Habitat, between two different Digital Grey Images (face of Einstein and a Map). Created with an Artificial Ant Colony, that uses images as Habitats, being sensible to their gray levels [in, V. Ramos, F. Almeida, “Artificial Ant Colonies in Digital Image Habitats – a mass behavior effect study on Pattern Recognition“, ANTS’00 Conf., Brussels, Belgium, 2000].
After “Einstein face” is injected as a substrate at t=0, 100 iterations occur. At this point you could recognize the face. Then, a new substrate (a new “environmental condition”) is imposed (Map image). The colony then adapts quickly to this new situation, losing their collective memory of past contours.
In white, the higher levels of pheromone (a chemical evaporative sugar substance used by swarms on their orientation trough out the trails). It’s exactly this artificial evaporation and the computational ant collective group synergy reallocating their upgrades of pheromone at interesting places, that allows for the emergence of adaptation and “perception” of new images. Only some of the 6000 iterations processed are represented. The system does not have any type of hierarchy, and ants communicate only in indirect forms, through out the successive alteration that they found on the Habitat. If you however, inject Einstein image again as a substrate, the whole ant society will converge again to it, but much faster than the first time, due to the residual memory distributed in the environment.
As a whole, the system is constantly trying to establish a proper compromise between memory (past solutions – via pheromone reinforcement) and novel ones in order to adapt (new conditions on the habitat, through pheromone evaporation). The right compromise, ables the system to tackle two contradictory situations: keeping some memory while learning something radically new. Antagonist features such as exploration and exploitation are tackled this way.
5 comments
Comments feed for this article
7 November, 2008 at 12:38 am
Gum voting « Chemoton § Vitorino Ramos’ research notebook
[…] we know that on the contrary, a delicate compromise between both is in fact of huge importance. The difficult but possible systemic trick is to remember the past as simultaneously innovating. If as a whole the system only remembers the past, no innovation is possible causing dramatic […]
29 May, 2009 at 2:20 am
Colour Image Segmentation using Evolutionary Computation « Chemoton § Vitorino Ramos’ research notebook
[…] this precise Map of Luanda, was used by me along with the face of Einstein to envision several dynamic image adaptive perception versus memory experiments trough ant-like artificial life systems…over what I then entitled – Digital Image […]
5 May, 2010 at 11:22 pm
Charles Darwin’s Scottish kilt « Chemoton § Vitorino Ramos' research notebook
[…] play Poker or the “Rock, Paper, Scissors” game. A typical confront between game memory (exploitation) and the ability to generalize (exploration). Where against precise opponents limited evolutionary […]
31 July, 2010 at 1:46 am
Metamorphosis « Chemoton § Vitorino Ramos' research notebook
[…] an activator-inhibitor system undergoes a change of parameters, one may pass from conditions under which a homogeneous […]
17 September, 2010 at 4:27 am
How to teach a computer on how to draw circles without teaching him anything about circles « Chemoton § Vitorino Ramos' research notebook
[…] at the title of this post, as many times you can, back and forth, in a non-explicit manner. Using stigmergy. You just implicitly create simple rules, even ant-like, non-anthropomorphic, which end-up at that […]