LHC physics dataset for unsupervised New Physics detection at 40 MHz

Area of Science:

• High Energy Physics
• Particle Physics
• Data Science

Background:

• The Large Hadron Collider (LHC) generates vast amounts of data from millions of proton-proton collisions per second.
• Real-time event selection systems in experiments like ATLAS and CMS are crucial for managing this data deluge.
• Current systems filter data to a manageable level, risking the loss of novel physics signals.

Purpose of the Study:

• Introduce a new dataset emulating real-time filtered collision events.
• Facilitate the development of novel event selection algorithms.
• Stimulate community-driven research in unsupervised new physics detection.

Main Methods:

• Creation of a curated dataset of proton collision events.
• Pre-filtering the dataset for events containing at least one electron or muon.
• Dataset designed to emulate real-time processing constraints (bandwidth, latency, computational resources).

Main Results:

• A valuable resource for testing and developing new event selection strategies.
• Enables assessment of sensitivity to new physics phenomena.
• Provides a benchmark for evaluating unsupervised learning algorithms.

Conclusions:

• The dataset supports the advancement of real-time data processing in particle physics.
• It is a key resource for exploring beyond the Standard Model physics.
• Encourages collaborative development of efficient and intelligent algorithms for future collider experiments.