Jul 6 – 13, 2022
Bologna, Italy
Europe/Rome timezone

Combining Conventional and Machine Learning Algorithms for LArTPC Reconstruction

Jul 8, 2022, 12:15 PM
Room 2 (Italia)

Room 2 (Italia)

Parallel Talk Neutrino Physics Neutrino Physics


Haiwang Yu (Brookhaven National Laboratory)


Liquid Argon time projection chamber or LArTPC is a scalable, tracking calorimeter that features rich event topology information. It provides the core detector technology for many current and next-gen large scale neutrino experiments, e.g., DUNE and the SBN program. For neutrino experiments, LArTPC faces many challenges in both hardware and software to achieve its optimum performance. On the software side, the main challenge is two-fold. First, deep domain knowledge needs further accumulation. Second, the event degree of freedom is high due to its large scale and uncertainties in the initial neutrino-argon interactions. With LArTPC R&D as one of its main goals, MicroBooNE has made major advancements in the LArTPC reconstruction paradigm building. Multiple fully-automated event reconstruction paradigms have been established. With the publishing of the initial results from a search for an electron neutrino low-energy anomaly, the effectiveness of these reconstruction paradigms are validated with real experiment data. This talk presents the Wire-Cell LArTPC reconstruction paradigm with particular highlights on how conventional and machine learning algorithms benefit from each other and fit into different tasks.

In-person participation No

Primary author

Haiwang Yu (Brookhaven National Laboratory)

Presentation materials