Noise Subtraction subgroup

• Pre-meeting
  • Date/Time: Tuesday, November 5th, 2024, 1600-1800 JST (0700-0900 UTC)
  • Connection Info.: Zoom1
  • Agenda

    • Brief Overview of Noise Subtraction Activities

      • Application to GW experiments/analysis
        • What can be accomplished? How effective is it in improving sensitivity/analysis etc.?

          • Example of NonSENS subtraction at H1, November 2nd, 2024: https://ldas-jobs.ligo-wa.caltech.edu/~detchar/summary/day/20241102/cal/nonsens/#gallery-9

      • Activities at KAGRA ( Who?, Project name?, **Features?(What can and what can't be done?)**, Current status?, Future plan?)

        • Algorithm
          • Liner ICA (Independent Component Analysis)

            • Yokoyama & Kume (RESCEU), Itoh & Kobayashi (OMU)

            • Ref.: https://doi.org/10.1093/ptep/ptaa056

            • ..
          • Non-liner ICA

            • Yokoyama & Kume (RESCEU)

          • DeepClean (CNN)

            • Yi, Chou, Yeh (NYCU, NTHU)

            • Ref.: https://journals.aps.org/prresearch/abstract/10.1103/PhysRevResearch.2.033066

            • Low-Latency and Offline Deployment

            • KAGRA F2F　Summer 2024 Slides: https://gwdoc.icrr.u-tokyo.ac.jp/cgi-bin/private/DocDB/ShowDocument?docid=15977

            • ..
          • LSTM
            • Takatani (OMU)
          • PINN (Physics-informed neural network)

            • DanChen & Sakemi

            • JPS 2023 Autumn JGW-G2315075

            • JPS 2024 Spring JGW-G2415603

          • Other ongoing initiatives?
            • Has each research unit or search pipeline been developing noise subtraction optimized for their specific needs?

        • Target noises & Witness channels

          • AC power 60 Hz and its harmonics
          • Acoustic noise (PSL, IFI, OMC, ...)
          • Other Dofs (CARM, MICH, PRCL)
          • ASC
          • Beam jitter (IMMT1 trans QPD)
        • Computers and Pipelines

          • GPU for offline DeepClean @ Taiwan

          • GPU for online DeepClean @ Kamioka

          • Kashiwa KMST
          • RESCEU
      • Other efforts within LV(K)
        • NonSENS (NON-Stationary Estimation of Noise Subtraction)

          • Derivation and description of algorithm: https://dcc.ligo.org/LIGO-T1800525

            • Estimates the optimal noise subtraction based on witnesses for the noise and witnesses for modulation of the noise coupling
            • Witness channel can be individual channels (linear subtraction), or product of two channels (bilinear subtraction)

          • Codebase: https://git.ligo.org/gabriele-vajente/nonsens

          • Running online at LHO/LLO
            • Front-end implementations:

              • LHO https://git.ligo.org/noisecleaning_o4/frontend-nonsens/lho-online-cleaning

              • LLO https://git.ligo.org/noisecleaning_o4/frontend-nonsens/llo-online-cleaning

            • The output channels are available in the control rooms, for use with commissioning, and are the same channels that analysis pipelines will use.
        • ...

    • Scope of the Group to be established

      • Organization?
        • Chia-Jui Chou from Natl. Yang Ming Chiao Tung U. will lead the group as the chief.

        • Will the group encompass all noise subtraction-related activities at KAGRA? (ICA+DeepClean+...)

        • Will the group focus specifically on areas that contribute to the commissioning timeline, i.e., low- or middle-latency?

        • or...
      • Roles/Goals for each phase
        • During the commissioning period prior to participating in O4b observation
        • During the O4b observation
          • Q: Will the cleaned h(t) produced by this team be used as an official deliverable and released to the public?!
        • For future phases O5/O6
      • How collaborate with on-site commissioners/IFO experts?
        • e.g.,

          • Commissioners may request specific noise subtraction, such as subtracting noise in channel-A using auxiliary channel-B?

          • Commissioners may manage a safe channel list?
          • ...
        • What considerations should guide us as we progress? How should we work with the team?
    • Safe channel, Witness channel, Auxiliary channel, ...
    • Online(front-end), Low-latency, Middle-latency, High-latency, (offline), ...