CIDeR-ML General Meeting

16 Apr 2026, 17:00 → 17 Apr 2026, 19:45 America/Los_Angeles

https://u-tokyo-ac-jp.zoom.us/j/83932834349

Minutes:

Quick recap

This meeting focused on planning and discussing workshop activities for the upcoming two-week CIDeR workshop. Team members presented their plans for various projects including cosmic muon reconstruction with Siren, flash matching using visibility and timing information, liquid argon TPC simulation comparisons, water transparency implementation in Lucid, PMT response modeling, and neutrino interaction framework development. Key discussions centered around technical approaches for implementing timing information in Siren models, computational efficiency concerns for flash matching algorithms, and the integration of different simulation components into existing frameworks. Several team members, including Zhenxiong, Sam, and Yifan, discussed their specific plans for cosmic calibration and reconstruction work, while others like Riya and the Lucid team outlined their goals for implementing new features and improving existing simulations. The conversation ended with a brief discussion of Mingzhuo's work on neutrino interaction modeling and generator comparisons, though some concerns were raised about the feasibility of certain approaches.

Next steps

• Zhenxiong: Complete the Cherenkov profile for cosmic muon and process the cosmic calibration with siren reconstruction information, aiming to finish preparation before the workshop; also process the reconstruction for cosmic muons with Siren pipeline and, if time allows, process training before the workshop.
• Zhenxiong: Investigate and, if necessary, reprocess the cosmic MC data to include timing information in the H5 files for Siren analysis.
• Zhenxiong: Start training sirenTV on the photon gun data as soon as the data loader for WCTE data format is available.
• Zhenxiong: Check offline with Patrick (and possibly Kaming) to obtain the path to the beam data and the reconstruction code for further analysis.
• Sam and Junjie: Work on extending flash matching to use visibility and timing in sirenTV, focusing on either full waveform-based optimization or extending the existing flash matching library to include waveform information, and evaluate computational efficiency and matching accuracy.
• Yifan and Pierre: Write up and document individual workshop plans (including simulation comparison and calibration fitting) in the shared Google Doc by Monday.
• Riya: Discuss with Omar on Monday the plan for one person to produce Michel electron simulation using Lucid and the other to work on integrating water transparency Z and radial dependence; start these tasks immediately and check in daily.
• Ryotaro: Verify LED photoemission distribution and generate MC for LED flash, then compare angular response from MC and real data; move the discussion on implementing MC to Lucid to the first week and start this discussion with Omar on Monday.
• Titouan and Nahuel: Redo plots of average charge per event per detector slice using iGPU, explore Lucid GitHub to find where attenuation length is used, test modifications, and clarify inputs/outputs for Siren; prepare and present a set of slides on Monday showing the proposed plan.
• Vera: Join Lucid explanations/hands-on sessions and work on implementing PMT response in Lucid; coordinate with Ryo for familiarization as needed.
• Mingzhuo: Fix coordinate transformation issue in model, use the same normalized coordinate for Siren input and loss calculation, and try different coordinates and numerical integration methods during the first week; email Lucas for suggestions and read relevant papers to inform model architecture.
• All participants with workshop plans: Prepare and present slides on Monday detailing their proposed tasks and timeline for the workshop.

Summary

Workshop Planning for Cosmic Muons

The team discussed workshop planning for the upcoming week, focusing on WCTE Photon Gun data and cosmic muon analysis. Zhenxiong presented a detailed plan to complete the Cherenkov profile for cosmic muons and process cosmic calibration with Siren reconstruction information. The team identified that timing information is available but not yet included in the H5 conversion.

sirenTV Data Implementation Discussion

The team discussed modifications to code for adding timing information and comparing Optic Siren with sirenTV. SLAC suggested starting sirenTV training on the photon gun data, which requires a data loader for WCTE format. Zhenxiong proposed comparing reconstructed data with frequent reconstructions and testing the impact of Siren reconstruction on cosmic calibration. Patrick suggested ultimately validating the approach using BEAM data to assess momentum reconstruction improvements. The team agreed to check offline about the location of the BEAM data.

sirenTV Application Workshop

SLAC discussed two potential applications for sirenTV during the workshop, with the main focus being on extending flash matching to use visibility and timing in large TPC applications. The proposed method involves using sirenTV to fit waveforms and optimize the position of particle clusters along the drift direction by comparing simulated PMT waveforms to ground truth data. SLAC outlined two approaches: a full reconstruction method and a stepwise method using photon library flash matching, with the latter being considered the easier implementation.

Computational Matching Approach Discussion

Kazuhiro and SLAC discussed the computational feasibility of implementing a new matching approach. They agreed that while the full n-squared combinations approach would be computationally expensive, using a two-step process with initial flash matching followed by waveform-level fine-tuning would be more practical. Kazuhiro suggested focusing the project on specific goals, either improving resolution through fine-tuning or addressing cases where flash matching fails, rather than implementing it as a general replacement method.

Simulation and Calibration Planning Update

Yifan outlined plans for the next two weeks focusing on two main areas: simulation comparison and calibration fitting. For the simulation comparison, Yifan will work on the reference simulation and stochastic version, while Pierre will focus on throughgoing muon cases. Regarding calibration fitting, Yifan plans to fit positions after making progress on fitting energy deposition sets, with the goal of improving input samples for calibration parameter fitting. Cesar will also work on a related project focusing on gradient correctness for uncertainty propagation.

Workshop and Project Planning Meeting

The team discussed plans for an upcoming workshop and two-week project timeline. Riya presented her plan to work on electron simulation and water transparency integration, while Patrick emphasized the need to decide on specific tasks (electron profile and water siren) by the end of the day to allow immediate start on Monday. The team also reviewed plans from other members including Vera's work on PMT response, Ryotaro's LED photoemission verification and MC generation, and Mingzhuo's neutrino interaction framework development. Patrick announced approval of a new 3-year US-Japan grant that will include support for neutrino interaction work.