stability homeostasis and learning
by Carson Reynolds
Following a conversation with Roz a few days ago I’ve been thinking more about notions of stability. Homeostatic behavior is pretty interesting; using notions of balance and stability as a way to force complex behavior offers up some cool possibilities, especially in light of models for classical mechanics.
I’m imagining things like a search system that keeps a notion of trajectory, and the inherent cost in altering search strategies. Basically, there is some sort of computational energy (viz. cycles) that is expended in optimization and search tasks. We need to model the cost of removing or changing constraints.