Neurotransmitter-Inspired Reinforcement Learning Modules for Neuromorphic Computing

Though they frequently lack the functional diversity of biological neural systems, neuromorphic computing systems have demonstrated promise in simulating brain-like processing. By incorporating five different neurotransmitter-inspired modules — dopamine, serotonin, acetylcholine, GABA, and glutamate — each mapped to particular cognitive processes and embodied reinforcement learning (RL) metaphors, this study offers a novel approach to neuromorphic chip design. We show how these modules can be combined to form a more adaptable and versatile neuromorphic architecture through the use of metaheuristic optimization techniques. We demonstrate our system through three experiments: (۱) parallel neurotransmitter processing improves multitasking performance, (۲) functional compensation when certain modules are compromised, and (۳) holistic integration enables emergent decision-making capabilities. The findings show that neuromorphic computing capabilities are greatly improved by biomimetic neurotransmitter diversity, especially in dynamic, complex environments that call for adaptive behavior.

کلیدواژه ها:

Neuromorphic Computing ، Biomimetic Neurotransmitter Modeling ، Embodied Reinforcement Learning ، Metaheuristic Optimization ، Cognitive Modularity