PhD position for anomaly detection with CMS

Building Robust and Calibrated Generative Models to Detect Anomalies in Data

Supervisors: Prof. Gregor Kasieczka (UHH), Prof. Timo Gerkmann (UHH)

Despite an impressive and extensive effort by the Large Hadron Collider (LHC) collaborations at CERN, currently, there is no convincing evidence for new particles produced in high-energy collisions. However, the Standard Model cannot be the final theory of nature. Past years have seen an enormous increase in anomaly-based strategies to search for new physics, such as the weakly supervised CATHODE approach co-developed in Hamburg. A key ingredient in this approach is training a generative model to learn an in-situ model of the background data.

This project will combine state-of-the-art techniques in quantifying the uncertainty of generative models and apply them to improve anomaly detection capabilities in particle physics to aid the potential discovery of new fundamental particles. To this end, data recorded by the CMS experiment will be analyzed.

Requirements:

• Master's degree in computer science or physics;
• proficient in Python and familiar with modern machine learning libraries;

Relevant expertise in at least one of these fields:

• particle physics data analysis/phenomenology and machine learning;
• deep flow/diffusion model experience/development;
• desirable: experience with generative machine learning models, anomaly detection techniques, or particle physics analysis pipelines.

Position:

• University of Hamburg;
• 75% EGR. 13 (TV-L) position for three years, pending approval of funding;
• 50% of the position will be funded by "Verbundprojekt 05H2024" (ErUM-FSP T03 - Run 3 von CMS am LHC: Elementarteilchenphysik mit dem CMS-Experiment).