Co-evolved SNN ensembles outperform traditional evolution via game theory

Evolutionary optimization of spiking neural networks (SNNs) becomes prohibitively difficult as task complexity grows because the combined topology-parameter space expands super-exponentially. To address this scaling challenge, researchers Catherine Rodriguez and James Ghawaly Jr. introduce a co-evolutionary ensemble framework where a population of candidate SNNs is evolved with fitness defined by each network's marginal contribution to group performance. Grounded in cooperative game theory and difference evaluation functions from multiagent systems, this credit assignment rewards networks that consistently improve ensemble performance and penalizes redundancy, encouraging complementary specialization during evolution rather than relying on post-hoc combination of independently trained networks.

The approach was evaluated on classification, regression, and control tasks under μCaspian neuromorphic hardware constraints. Co-evolved ensembles achieved statistically significant improvements over both single-network evolution and post-hoc ensembles across all tasks. The most pronounced gains appeared in control tasks, where standard evolution failed to discover effective policies and co-evolution enabled a qualitative transition to near-optimal performance. Accepted at the 2026 ACM International Conference on Neuromorphic Systems (ACM ICONS 2026), this work demonstrates a scalable path for neuromorphic AI that could unlock complex real-time control applications.