Towards Neuromorphic Evolutionary Computation: A Position Paper
IEEE World Congress on Computational Intelligence (WCCI) 2026, Maastricht, Netherlands
Abstract
Neuromorphic Computing (NC) and Evolutionary Computation (EC) are mature, yet largely fragmented. This position paper argues that their convergence is timely and technically plausible. We define Neuromorphic Evolutionary Computation (NEC) as a paradigm that bridges NC and EC. We review the current landscape and its limitations, with an emphasis on the prevalence of hybrid execution and incomplete energy-latency reporting. We then present a probable Cortico-Basal-Thalamic Loop (CBTL)-inspired conceptual architecture for operator selection and outline what must be implemented on-chip to achieve an NEC approach, including operator embodiment, associative memory, and state estimation. We summarise open challenges spanning numerical precision, scalability, learning rules, benchmarking, theory, and tooling, including the need for convergence analysis framed in terms of the best-so-far process. We close with a phased roadmap and concrete reporting requirements towards a reproducible NEC ecosystem.