Exploiting Activity for the Modeling and Simulation of Dynamics and Learning Processes in Hierarchical (Neurocognitive) Systems

摘要

Although modeling and simulation depend on each other, there is no means to formally simplify models and corresponding simulations at the same time. The activity concept elicits the coordination and the number of computations of a system, highlighting salient features about its dynamics. I follow here a neurocognitive example linking and applying definitions and algorithms based on an (neuronal) activity measure. At the modeling level, activity state regions are identified dynamically. At the simulation level, I present how to track the activity region at the component level. At the learning level, I finally present an activity-based search algorithm that is able to find the best components (actions) in a network (a series of actions). Activity regions are used hierarchically from neurons to actions.