Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-11553

Investigate drifting positions

    Details

    • Type: Story
    • Status: To Do
    • Resolution: Unresolved
    • Fix Version/s: None
    • Component/s: meas_deblender
    • Labels:
      None
    • Story Points:
      3
    • Sprint:
      DRP F17-4, DRP F17-5, DRP F17-6, DRP S18-1, DRP S18-2, DRP S18-3, DRP S18-4, DRP S18-5, DRP S18-6, DRP F18-1
    • Team:
      Data Release Production

      Description

      Some sources consistently want to drift from their initial position to the same (incorrect) location in the deblender. The attachments show one example, where a peak (peak 4) that is mostly outside the footprint wants to move in the direction of the bright star next to it.

      This behavior prevents convergence, since the algorithm is stuck resetting the position every 100 iterations. This ticket is to search for a better method to identify and handle similar situations.

        Attachments

          Issue Links

            Activity

            Hide
            fred3m Fred Moolekamp added a comment -

            Some progress was made several months ago before we began testing the deblender on HSC data. Peter Melchior has a solution in his work on a new deblender branch. Testing (DM-11909) will determine if there is still a problem, otherwise this ticket can be closed (I adjusted the number of story points for work already completed).

            Show
            fred3m Fred Moolekamp added a comment - Some progress was made several months ago before we began testing the deblender on HSC data. Peter Melchior has a solution in his work on a new deblender branch. Testing ( DM-11909 ) will determine if there is still a problem, otherwise this ticket can be closed (I adjusted the number of story points for work already completed).
            Hide
            fred3m Fred Moolekamp added a comment -

            We believe that the drifting sources are caused by bad difference images for fainter sources. The algorithm that updates the positions assumes that there is a dipole in the residual of data-model for each source and fits for the position that minimizes the dipole. For sources next to much brighter objects there is no dipole to fit, which not only causes drifting but can also result in completely non-sensical position updates that cause scarlet to crash.

            Show
            fred3m Fred Moolekamp added a comment - We believe that the drifting sources are caused by bad difference images for fainter sources. The algorithm that updates the positions assumes that there is a dipole in the residual of data-model for each source and fits for the position that minimizes the dipole. For sources next to much brighter objects there is no dipole to fit, which not only causes drifting but can also result in completely non-sensical position updates that cause scarlet to crash.

              People

              • Assignee:
                fred3m Fred Moolekamp
                Reporter:
                fred3m Fred Moolekamp
                Watchers:
                Fred Moolekamp, Peter Melchior
              • Votes:
                0 Vote for this issue
                Watchers:
                2 Start watching this issue

                Dates

                • Created:
                  Updated: