Speaker
Manqi Ruan
Description
Based on ParticleNet, ParticleTransformer, and physics prior knowledge, we modified the model architecture and achieved a significant improvement in performance on the 2024 experimental data of the Crab Nebula. In addition, we explored a semi-supervised learning approach. Using the SimSiam training framework, we performed extensive pretraining on a large amount of experimental data, which helped alleviate the performance degradation caused by the mismatch between Monte Carlo simulations and real experimental data.
Primary authors
Chao Yang
(IHEP)
Manqi Ruan
永峰 朱
