Bit-vector optimization algorithms for control of learning in neurons with second-messenger dynamics

The authors describe comparatively simple; evolutionary algorithms that can utilize point mutation or crossover operators or both. The algorithms are first tested on two predefined functions of excitase configurations, and their basic strengths and weaknesses are revealed. They are then tested in vivo, coupled with the reaction-diffusion dynamics of the cell to execute a particular task (here, the motion of a robot on a plane). This provides insight into the interaction between the fast dynamics of cellular processing, and the choice of slow dynamics, or motion through the parameter spaces of learning.<>