Co-Evolved Spiking Neural Network Ensembles via Marginal Contribution Fitness
Researchers have developed a novel co-evolutionary framework for optimizing spiking neural networks (SNNs), addressing the challenge of scaling with task complexity. This approach, inspired by cooperative game theory, defines fitness based on each network's marginal contribution to the ensemble's performance, encouraging specialization and complementarity. Evaluations on classification, regression, and control tasks under neuromorphic hardware constraints demonstrated significant improvements over single-network evolution and post-hoc ensembles, particularly in complex control scenarios where standard methods failed. AI