Göttingen 2025 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 76: Data, AI, Computing, Electronics VII (Generative AI, MC Generators)

T 76.3: Vortrag

Donnerstag, 3. April 2025, 16:45–17:00, VG 2.101

Study of deep generative models for the enhancement of simulated ATLAS datasets — Boris Flach, Andre Sopczak, and •Lukas Vicenik — Czech Technical University in Prague

Numerous searches for new particles and precision measurements crucially depend on the amount of available simulated data, which has an impact on the resulting analysis uncertainties. For instance, machine learning algorithms for separating signal and background events could significantly profit from enlarged simulated datasets. We propose advanced generative models based on variational autoencoders, generative adversarial networks, and diffusion-based deep generative models to address the limitations of current simulated datasets. These models generate synthetic data that capture complex, non-homogeneous features observed in particle physics. Evaluation metrics from particle physics and machine learning are employed to assess the accuracy, diversity, and physical validity of the generated data. The augmented datasets are subsequently used to enhance signal and background separation, reduce uncertainties in analyses, and improve the overall reliability of the results.

Keywords: generative models; data augmentation; ATLAS; CERN