ILC-Japan physics seminar #3: Machine learning on event reconstrruction

Monday 31 Mar 2025, 14:00 → 15:05 Asia/Tokyo

ILCやヒッグスファクトリーの物理を主なトピックとして、ILC-Japan physics seminarをZoomにて定期開催します。
第三回のテーマはMachine learning on event reconstructionです。講演は英語です。日本語の質問も受け付けます。

We (as ILC-Japan physics WG) are holding a series of seminars related to physics on ILC and Higgs factories. The topic of the third seminar is machine learning on event reconstruction. Talks will be given in English.

Zoom connection info:

https://u-tokyo-ac-jp.zoom.us/j/84554632255?pwd=ULm1jI5ElV05sHzpTw1NWeDhHZuFD1.1

meeting ID: 845 5463 2255
passcode: 843129

14:00 → 14:05 News on ILC/Higgs factories (if any)¶

14:05 → 15:05 AI-based event reconstruction and detector design for Higgs factories¶

Detector designs for Higgs factories are highly optimized to maximize performance of multi-particle final states, especially including multiple hadron jets. This requires highly-granular detector elements, especially in calorimeters to separate particles. Event reconstruction with such detectors is highly non-trivial, requiring intelligent pattern recognition, thus suitable to use modern machine-learning techniques. In this talk I will introduce several state-of-the-art technologies such as transformer-based flavor tagging, GNN (graph neural network) based particle flow and so on. These algorithms are also crucial for detector design, since performance of the detectors is highly correlated to the reconstruction algorithms and ML-based algorithm has advantages on utilizing additional or improved information of improved detector designs seemlessly to compare with existing designs. Discussions on such direction will also be presented.

Speaker: Taikan Suehara (ICEPP, The University of Tokyo)

250331-hfreco-ilcjphysseminar-suehara.pdf