---

Why is This an Important Research Topic?

Nengo provides a common abstraction for developing SNNs using neurons and ensembles. Mapping new backends enables us to use the same codebase to target new and novel architectures.

Current Status

This project was initiated by Terry Stewart and others at the 2021 Telluride workshop.

[Relevant repo from Terry Stewart](https://github.com/neuromorphs/ant-nengo-backends]

Neuron models

• Creating the neuron and the I/O flow are the key pieces of developing a backend.
• Could write a simple neuron model in C code
  • Rectified linear neurons as a starting point
• FPAA model could interface with Scilab/XCOS
  • Does the IO happen all at once (batch) or does it happen in a feedback loop type format?
  • An online, real-time closed-loop system
  • Can pass input data to TextAPI but it would be more offline and batch
  • Loihi and BrainDrop can be operated in “batch” mode with precomputing the inputs