SMARTHEP NGT Edge ML School

I was delighted to be involved in the planning of the SMARTHEP Edge ML School, which was hosted at CERN from 23-27 September 2024. We organised a program of lectures and hands on tutorials on topics relating to Edge ML including:

  • Model compression techniques including quantization aware training
  • Efficient GPU inference
  • Efficient FPGA inference
  • Neuromorphic computing and spiking neural networks
  • Heterogeneous accelerated computing
  • Geometric deep learning
  • Realtime applications in earth-monitoring satellites
  • Ultra low power AI processors
  • Inference at the speed of data

Take a look at the event page for reading material: https://indico.cern.ch/e/SMARTHEP-edge-ML

The agenda gave participants a broad overview of aspects relating to Edge ML, from the ML theory side to the practical CS and Engineering deployment side. We were pleased to gather expert speakers from diverse fields, giving unique perspectives to participants who will go on to make their contributions to, and uses of Edge ML. I presented a short tutorial on using conifer to run inference of Decision Forests on FPGAs with low latency and high throughput, including a demo of the new Forest Processing Unit, that you can find here.

We included a poster session for participants to present their own work, and awarded prizes for the best posters as chosen by a jury of the organisers. Congratulations to Noah Clarke Hall, Denis-Patrick Odagiu, and Simone Machetti for their work!

SMARTHEP NGT Edge ML School Poster Awardees

I thank my fellow organisers for making this event a great success: Anna Sfyrla, Maurizio Pierini, and Thea Aarrestad.