Choose timezone
Your profile timezone:
CIDeR-ML General Meeting
→
America/Los_Angeles
Description
https://u-tokyo-ac-jp.zoom.us/j/83932834349
Recording
Minutes:
Quick recap
The meeting focused on planning for an upcoming workshop in Japan, where the team discussed implementing SirenTV and related simulations. Junjie presented a plan to help the Japan team with SirenTV implementation, including scaling studies using LUCiD and updating the ROOT-to-H5 pipeline for timing information. The team discussed the need for collaboration between US and Japan sides, with specific tasks assigned including water transparency position implementation and calibration demonstrations. Zhenxiong reported on validation studies of OpticSIREN performance, finding biases in charge measurements and concluding that sampling parameters don't significantly impact results. The team also discussed MDT (PMT response and electronics modeling) development, with Ryo and Ryotaro planning to work on implementing angular response functions based on cosmic analysis data during the workshop period.
Next steps
- Junjie: Work on updating the ROOT-to-H5 pipeline for SirenTV, including adding timing information, starting this week and continuing during the first week in Japan.
- Junjie: Transfer knowledge on the updated ROOT-to-H5 pipeline to a designated person from the Japan side for future maintenance and use.
- Riya, Nahuel, and Titouan: Continue work on implementing water transparency (or other analytical forms) into the simulation, with the goal of demonstrating calibration ability during the workshop.
- Junjie: During the first week in Japan, start training and validation on SirenTV using available datasets, and verify performance against other models and MC2s.
- Ryo/Ryotaro: Develop and implement the angular response function from cosmic analysis into MDT, and validate it using LED simulation data during the upcoming workshop period.
- Sam: Help Riya set up and integrate the SLAC cluster with VS Code, ensuring Riya is up and running before the workshop.
- Patrick and Omar: Check on the readiness and status of the LUCiD IWCD production dataset for scalability studies, and coordinate on generating or providing the necessary samples.
- Ryotaro: Continue single PE analysis and discuss with Laurence, Jacob, and others on integrating charge information for calibration and single PE estimation.
- Zhenxiong: Investigate the Cherenkov profile to improve OpticSIREN simulation results, and test the updated reconstruction pipeline with cosmic calibration information before the workshop.
- Zhenxiong: Visualize and compare the before/after tuning of the SIREN model in a meaningful way (e.g., integrating/aggregating/slicing data) to identify which parts of the detector are most affected by tuning.
Summary
SirenTV Implementation in Japan Plans
Junjie presented plans for implementing SirenTV in Japan, outlining two main tasks: creating a larger (IWCD) dataset using LUCiD to test scalability (up to tens of thousands of PMTs) and updating the ROOT-to-H5 pipeline to include timing information. The team discussed conducting training and validation during the first week in Japan, with Junjie noting the need for collaboration with the Japan side, particularly regarding calibration implementation. The second week would focus on demonstrating SirenTV's ability to change outputs with new simulations and water transparency calculations, though Junjie acknowledged that many tasks might spill into the second week due to ongoing preparation work.
Riya announced she is working on water transparency projects with two French interns, Nahuel and Titouan, who are currently getting familiar with LUCiD. Patrick discussed the need for personpower in Japan, noting that while immediate leadership on this topic is challenging due to personpower issues, the workshop will serve as a seed study for future work. Junjie and Patrick discussed the conversion process, with Junjie planning to use available GPUs for validation and production, though he clarified that the conversion is typically done on CPU. Sam agreed to help Riya with integrating the SLAC cluster with VS Code.
OpticSIREN Simulation Performance Updates
Zhenxiong presented updates on validating the performance of OpticSIREN simulation by comparing PMT responses using different simulation processes and inputs. She found that changes in sampling parameters, including step sizes and file numbers, did not improve results, indicating that sampling settings are not the cause of observed biases or uncertainties. Zhenxiong plans to investigate the Cherenkov profile and test the updated reconstruction pipeline with cosmic calibration. Patrick Tsang suggested visualizing how the model changes before and after tuning to better understand which detector parts are most affected.
MDT Development and Implementation Planning
Japan team discussed the purpose and implementation of MDT (PMT response and electronics modeling), which can be used separately from the main WCSim simulation to allow for quicker changes to PMT response modeling without rerunning the entire simulation. Ryo agreed to develop MDT during the upcoming workshop period, starting with implementing angular response functions based on cosmic analysis data. The team also briefly discussed Ryotaro's ongoing single PE analysis work, which is progressing well with charge distribution studies, and considered potential student involvement in future development work for both MDT and fiTQun.
