Fast Machine Learning for Science Conference 2026

Real-time Event-level Zero-shot Learning at 40 MHz

Not scheduled

12m

Presentation Contributed Talks

Speaker

Leo Rozanov

Description

Zero-shot learning (ZSL) refers to the ability of a model to be able to classify unseen labels. We present High Granularity Quantization (HGQ) Linformer based event-level ZSL architecture model trained using contrastive methods on the Collide-2V dataset. The output of the model is an embedding vector which enables classification through ZSL. Different physics processes occupy unique regions in the embedding space which also allows unseen processes to be distinguished. We evaluate the performance with both seen standard model processes and unseen processes for both standard model and beyond standard model, achieving good performance. The architecture serves as a proof-of-concept for future direction for both trigger and scouting methods for the LHC.