Scalability analysis of the event-driven scheduler

Establish complexity bounds for the priority queue used by the event-driven scheduler for networks of Petri neurons and explore distributed execution strategies for deploying the scheduler on multi-core or FPGA-based neuromorphic platforms.

Background

The paper introduces an event-driven scheduler based on a priority queue to execute T-timed Petri-net representations of neural circuits. Although the scheduler is validated on three small microcircuits, its behavior and computational cost for large neural networks are not analyzed. The unresolved issue is whether the priority-queue implementation can scale efficiently as the number of neurons and scheduled events increases, and how execution could be distributed across multi-core processors or FPGA-based neuromorphic platforms. The authors identify the derivation of complexity bounds and the investigation of distributed execution as concrete future work.