sahooroshni/semester-thesis-ethz

Semester thesis titled "Top jet training region for Normalized Autoencoders in the t-channel Semivisible Jet Analysis in CMS" at ETH Zürich

Semester Thesis at ETHZ

Abstract: The existence of Dark Matter (DM) provides strong evidence for physics beyond the Standard Model. The DM resides within a ‘dark sector’, which, if it is a strongly coupled sector, can result in jet-like collider objects in the Large Hadron Collider called ‘semivisible’ jets. Semivisible jets are collimated sprays of invisible dark hadrons and Standard Model hadrons. Normalized Autoencoders (NAE) can be used as anomaly detection algorithms to perform signal-agnostic searches for semivisible jets. This can be done by training the NAE to recognize ordinary Standard Model (SM) jets to tag potential semivisible jets as ‘anomalous’ objects. We designed a signal-free training region composed of top-quark jets arising from tt¯ events and demonstrated that this is a suitable region to train the NAE directly on real data in order to classify semivisible jets from Standard Model top-quark jets.

This project was done in the group of Prof. Dr. Annapaola de Cosa in the Institute of Particle Physics and Astrophysics (IPA) at ETH Zürich, under the supervision of Florian Eble and Dr. Roberto Seidita.