This tutorial presents an in-depth introduction to YANA, a many-core near-memory-computing Spiking Neural Network (SNN) FPGA accelerator developed at the FZI Research Center for Information Technology. The accelerator targets embedded sensor processing applications in medical, industrial, and automotive contexts, with a focus on dataset evaluation and real-time processing of high data rate neuromorphic sensors. Participants will gain hands-on experience with key co-design decisions such as quantization, the mapping of logical neurons onto physical processing elements, and the accelerator's integration within a System-on-Chip (SoC) FPGA environment running Linux. The YANA framework will be made available open source ahead of the tutorial.

Tutorial Outline (preliminary)

1. Introduction, Overview & Setup
2. On participant laptops:
   • Train & evaluate a quantized deep SNN using Norse, applying different levels of pruning
   • Export the trained model using NIR for further processing
   • Map and partition the deep SNN onto the YANA many-core accelerator using the YANA toolchain
     • Evaluate results for different parameterizations
   • Transfer generated accelerator init files to the KR260 platform
3. On KR260:
   • Initialize the YANA accelerator
   • Evaluate test samples
   • Evaluate accelerator performance across different pruning levels

Contact

Brian Pachideh — Cc: Sven Nitzsche, Moritz Neher

Further announcements — Additional information will be published on this page as the conference approaches. This includes preparation steps and any software participants should have ready on their laptops before the tutorial, downloadable resources shared ahead of time, and details on what to bring on the day. Please check back closer to the event.

---

YANA found its HEART in Heidelberg.