Wednesday, 6 July 2016

ALife day 3

Day 3 of ALife, and the great science continues!

First up was the keynote by Jorge M. Pacheco, on “Linking Individual to Collective Behavior in Complex Adaptive Networks”. A nice discussion of investigating iterated prisoner-dilemma cooperation-defection situations where the agents are distributed over social networks. Agents can change their behaviours (from Cooperator to Defector or v.v, by copying the strategy of their most successful neighbours), or change their network (both Cs and Ds want to drop connections to Ds, but Ds want to make connections to Cs). Interestingly, it seems that changing the network (changing who your friends are) is more effective than changing your strategy (changing what you do). So it’s best to isolate defectors. (Hmm.)

Next was the morphology session. Although several of the speakers admitted their work wasn’t truly about morphology, all the talks were all interesting. We heard about difficulties of co-evolving morphology and body controllers. Morphology seems to converge quickly, because if it changes, the co-evolving brain can’t adapt fast enough. The speaker had some suggestions on how to improve the situation. Next we heard about evolving soft body robots, exploiting “passive dynamics” and using this capability as a sort of embodied “computational reservoir”. Then there was an examination of how the shape of space (a “donut” shaped torus v a “bicycle tyre” shaped torus) affects iterated prisoner dilemma: donuts are better. Then finally there was a nice talk about co-evolving predator field of view and prey “swarminess”, with interesting Red Queen style oscillations: prey evolve to swarm to confuse the predator, which evolves a narrower field of view to avoid confusion, so the prey then evolve to scatter to hide from the focussed predator, which evolves a wider field of view, and so on.

The second keynote of the day was Mark Bickhard, talking on “Cognition and the Brain”. In a brilliant talk, he covered what it means to be an anticipatory system (having a set of possible future actions to choose from), and how that can allow representation to be true or false (to see if a representation is true, wait and see what the future brings). The talk wove together this philosophy with details about microstructures in the brain, particularly the possible role of glial cells being large scale slow processes that modify the attractor landscape of the brain to influence smaller scale faster neural processes. The whole approach sits within a process philosophy, which permits the emergence that duality is designed to make impossible. I now want to go away and simulate the nested oscillator modulatable resonant architecture he speculated might underlie these processes.

The final session of the day was on computational biology, with a range of talks covering the self-organisation of badger latrines, chopping the tails off tadpoles, making C.Elegans models that swim correctly, a multiscale simulation of E.Coli (from molecular to Petri dish scales), and experiments on the evolution of genetic networks. That’s quite a varied bunch!

No comments:

Post a Comment