FNS

FNS implements event-driven simulations of spiking neural networks (SNNs) using the Leaky Integrate-and-Fire with Latency (LIFL) neuron model to enable efficient, low-energy simulation of realistic neuronal dynamics for applications such as implanted neuroprostheses.

Key Features:

• Event-driven Approach: Employs an event-driven methodology that updates neuron states only upon spike events to focus computation on active events.
• LIFL-based Model: Uses the Leaky Integrate-and-Fire with Latency (LIFL) spiking neuron model to capture essential neuronal dynamics with reduced resource requirements.
• Heterogeneous Neuron Groups and Multi-scale Connectivity: Supports heterogeneous neuron groups with multi-scale connectivity, including delayed connections and plastic synapses.
• Parallelization Strategy: Implements a novel parallelization strategy to enable precise multi-threaded simulations on limited hardware configurations.
• Periodic Dumping Mechanism: Includes a periodic dumping mechanism for efficient management of simulation data over extended time scales.

Scientific Applications:

• Population and network-level studies: Exploring interactions within and between populations of spiking neurons.
• Long-term simulations on constrained resources: Enabling extended-duration simulations while minimizing memory and computation demands.
• Low-energy brain modeling and implanted neuroprostheses: Supporting brain modeling use cases where reduced energy consumption is required, including implanted neuroprostheses.

Methodology:

Combines the LIFL neuron model with an event-driven simulation approach, supports heterogeneous neuron groups and multi-scale connectivity (including delayed connections and plastic synapses), incorporates a novel parallelization strategy for multi-threaded execution and a periodic dumping mechanism, and has been evaluated against similar models in NEST for simulation time and memory usage.