Neurological Based Timing Mechanism for Reinforcement Learning

The inherently time-dependent dynamics which underly the neuronal spiking communication, are ubquitous throughout brain, and yet are not fully understood. Likewise time-based mechanisms are underdeveloped in the field of Machine and Reinforcement Learning (RL) [7]. The complexity-rich and multi-dimensional dynamics observed in the brain offer potential advancements in Machine Learning (ML), and development of Artificial Generalized Intelligence. It is in our interests to model known time-mechanisms of neuronal spiking communication, and reproduce the emergent properties of complex timing and learning in assemblies. If neuronal temporal dynamics can be understood, a new field of possibilities will open for in-situ models which learn in complex real-time environments. A key challenge for these models is correctly identifying associations of actions and stimulus at variable time separations. In this article, we bring the flexible time representation mechanisms from neuroscience to the field of automata and RL, to explore its potential.