Tutorial: Neuromorphic Evolutionary Computation (NEC)

IEEE WCCI/CEC, Maastricht, the Netherlands

Neuromorphic Computing Computational Intelligence Spiking Neural Networks Evolutionary Computing Metaheuristics Event-driven Algorithms Low-power Embedded Systems

A tutorial on neuromorphic-embedded computation, guiding participants through spike-driven optimisation frameworks, hands-on use of neuromorphic tools and practical implementations in event-driven environments.

Abstract

Neuromorphic computing brings spiking dynamics and event-driven efficiency into the realm of optimisation. Evolutionary algorithms, long valued for their versatility, can now be re-imagined on this substrate. We first present the theoretical principles behind neuromorphic-based metaheuristics, highlighting their motivations, classification, and trade-offs. We then detail two neuromorphic evolutionary computation approaches: one includes a practical architecture in which units implement mutation and variation via spikes, and the other leverages the Basal-Ganglia-Thalamic loop. Finally, we discuss practical applications using open-source packages, showing participants how to design, run, and interpret neuromorphic optimisation experiments. This tutorial aims to demonstrate how spiking computation can shape a new generation of evolutionary search and open the way to applications in robotics, IoT, and embedded intelligence.

Learning Objectives

The main objective of this session is to equip participants with both conceptual and practical tools to understand and apply neuromorphic-based evolutionary computation algorithms.

In particular, by the end of the tutorial, participants will:

Understand the theoretical foundations of neuromorphic evolutionary optimisation.
Recognise the state-of-the-art architectures and their trade-off.
Learn from the authors the practical details and rationale for using and modifying the Neuroptimiser package to run experiments.
Identify research opportunities for scaling, hybridisation, and applications in WCCI’s fields.

Tutorial Outline

Introduction and Foundations (30 min)
- Metaheuristics and evolutionary computation basics
- Neuromorphic computing and spiking dynamics
- Theoretical framework of neuromorphic metaheuristics
Practical Framework: The NeurOptimiser (40 min)
- Neuromorphic Heuristic Units and Spike-driven mutation strategies
- Hands-on exercises on benchmark functions from zero to hero
- Feasible adaptations and tweaks on the framework
Perspectives and Future Directions (20 min)