Optimizing the Efficiency of Winner-Takes-All Neuromorphic Circuit Optimization Using Self-Adaptive Multi-Population Quadratic Approximation Guided Jaya Algorithm

Metaheuristics are robust and sophisticated approaches to solving Electronic Design Optimization problems. However, due to the non-linearity of these optimization problems, the complexity increases and many of these algorithms do not deliver the global optimum. Additional difficulties include diverse constraints, inherent errors, conflicting objectives, and multiple local optima. Consequently, significant variations in the final results of these problems could be observed across multiple iterations, even after using traditional meta-heuristics. Therefore, proper tuning of the control parameters of these algorithms is very important, since it is proportional to their numerical cost and accuracy. The primary objective of this investigation is to enhance both the stability and quality of outcomes while optimizing the Winner-Takes-All neuromorphic circuit using a recently proposed parameter-free approach called Self-adaptive multi-population Quadratic Approximation guided Jaya algorithm. Extensive experimentations with promising outcomes confirm its efficiency compared to other state-of-the-art counterparts. Finally, validation is performed using the circuit design tool Cadence Virtuoso, further illustrating a close agreement with the algorithmic results.