Upgrade to Pro — share decks privately, control downloads, hide ads and more …

GT ICR - Antoine Marchal - IViS: Towards Green,...

GT ICR - Antoine Marchal - IViS: Towards Green, Scalable Imaging of diffuse emission lines for the SKA Era

Avatar for François Orieux

François Orieux

February 16, 2026
Tweet

More Decks by François Orieux

Other Decks in Research

Transcript

  1. IViS: Towards Green, Scalable Imaging of diffuse emission lines for

    the SKA Era ! Antoine Marchal - CRCN CNRS @LPENS GT-ICRÑFeb 2026 Processing the GASKAP-HI full-survey w/ the GASKAP-HI collaboration imaging team Hiep Nguyen, David McConnell, James Dempsey, Naomi McClure-GrifÞths, Nick Pingel, Callum Lynn, Jackie Ma, Jo Dawson
  2. 2 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Synthesis

    imaging with mosaicking/joint deconvolution using a non-linear optimization approach Today..
  3. 3 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Imaging of Òlarge objectsÓ (Cornwell 1988, Sault et al. 1996) Beams
  4. 4 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Imaging of Òlarge objectsÓ (Cornwell 1988, Sault et al. 1996) Beams Visibilities Subset of the FT
  5. 5 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Imaging of Òlarge objectsÓ (Cornwell 1988, Sault et al. 1996) Beams Visibilities Subset of the FT
  6. 6 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Imaging of Òlarge objectsÓ (Cornwell 1988, Sault et al. 1996) Beams Visibilities Subset of the FT
  7. 7 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Imaging of Òlarge objectsÓ (Cornwell 1988, Sault et al. 1996) Beams Visibilities Subset of the FT Joint Deconvolution
  8. 8 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Imaging of Òlarge objectsÓ (Cornwell 1988, Sault et al. 1996) Beams Visibilities Subset of the FT Joint Deconvolution How do we do that?
  9. 9 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    A brief (& biased) history of HI surveys (in the local Universe) Instruments Sorted by end-date of observations
  10. 10 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    A brief (& biased) history of HI surveys (in the local Universe) Imaging methods Sorted by end-date of observations
  11. 11 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Multi-scale CLEAN vs MEM ? Multi-scale CLEAN Maximum entropy method For HI, see Stanimirović et al 2002 Cornwell et al 2008 See review by Narayan & Nityananda 1986 1986 ÒParametricÓ multi-Gaussian model Multi-scale CLEAN vs MEM ?
  12. 12 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Multi-scale CLEAN vs MEM ? For HI, see Stanimirović et al 2002 Cornwell et al 2008 See review by Narayan & Nityananda 1986 1986 Multi-scale CLEAN Maximum entropy method ÒParametricÓ multi-Gaussian model Multi-scale CLEAN vs MEM ?
  13. 13 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Multi-scale CLEAN vs MEM ? || ˜ − ||2 Σ − 1 ζ ∑ i Ii ln ( Ii Mi ) For HI, see Stanimirović et al 2002 Cornwell et al 2008 See review by Narayan & Nityananda 1986 1986 Multi-scale CLEAN Maximum entropy method ÒParametricÓ multi-Gaussian model Non -parametric model Multi-scale CLEAN vs MEM ?
  14. 14 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    Multi-scale CLEAN vs MEM ? || ˜ − ||2 Σ − 1 ζ ∑ i Ii ln ( Ii Mi ) For HI, see Stanimirović et al 2002 Cornwell et al 2008 See review by Narayan & Nityananda 1986 1986 1- Positivity 2- Smoothness ÒParametricÓ multi-Gaussian model Non -parametric model Multi-scale CLEAN Maximum entropy method
  15. 15 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    A brief (& biased) history of HI surveys (in the local Universe) Imaging methods Sorted by end-date of observations
  16. 16 08/12/2025 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging

    A brief (& biased) history of HI surveys (in the local Universe) Imaging methods Sorted by end-date of observations ?
  17. 17 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging A

    brief (& biased) history of HI surveys (in the local Universe) Australian Square Kilometer PathÞnder ASKAP antennas PAF - 36 antennas (12m diam. each) - Freq. range 700 MHz - 1.8 GHz - Bandwidth 300 MHz - FoV ~30 square degrees - Baseline up to 6 km - 8 Survey Science Projects A B C ASKAP: a survey machine
  18. 18 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging A

    brief (& biased) history of HI surveys (in the local Universe) ASKAP: a survey machine GASKAP-HI survey with the ASKAP 4000h of observations 50h / field 30h / field 200h / field
  19. 19 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging What

    are the challenges ? 2- Image quality and fidelity 1- Deep integration ! 3- Green and Scalable ? A brief (& biased) history of HI surveys (in the local Universe)
  20. 20 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging What

    are the challenges ? 2- Image quality and fidelity 1- Deep integration ! 3- Green and Scalable ? A brief (& biased) history of HI surveys (in the local Universe)
  21. 21 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging IViS

    (Interferometric Visibility-domain Inversion Software) Re-inventing the wheel: A (not so new) non-linear optimization approach to join deconvolution
  22. 22 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution Parkes telescope An open source code Ñ written in PyTorch framework (with ReadtheDocs, notebooks, É)
  23. 23 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution Parkes telescope A modular framework
  24. 24 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution 108 beams I(r) ˜ k (I′ k ; u, v) ≃ Beams
  25. 25 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution 108 beams Projected sky image I(r) ˜ k (I′ k ; u, v) ≃ I′ k (ℓ, m) Beams
  26. 26 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution 108 beams Projected sky image I(r) ˜ k (I′ k ; u, v) ≃ I′ k (ℓ, m) × Ak (ℓ, m) Primary beam Beams
  27. 27 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution 108 beams Projected sky image I(r) ˜ k (I′ k ; u, v) ≃ ℱ[ I′ k (ℓ, m) × Ak (ℓ, m)] Primary beam NuFFT (GPU) Beams similar to MPol (by Ian Czekala, see Zawadzki et al. 2023) Beams
  28. 28 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution L1,k (I′ k ) = ˜ k (I′ k ) − k Jk (I′ k ) = 1 2 ∑ u,v ( L1,k(I′ k) Σ1,k ) 2 108 residual Q(I′ k ) = 108 ∑ k Jk (I′ k ) + R(I(r)) Spatial regularisation Joint deconvolution with spatial regularisation One cost function 108 chi square
  29. 29 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution Parkes telescope Visualising the optimization at one frequency
  30. 30 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution Visualising A PPV cube GASKAP-HI survey with the ASKAP
  31. 31 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution Visualising A PPV cube GASKAP-HI survey with the ASKAP 4000h of observations
  32. 32 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution GASKAP-HI survey with the ASKAP 4000h of observations Visualising A PPV cube
  33. 33 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Re-inventing

    the wheel: A (not so new) non-linear optimization approach to join deconvolution GASKAP-HI survey with the ASKAP 4000h of observations Visualising A PPV cube
  34. 35 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT CASA 10h observation with MeerKAT Measurement set
  35. 36 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT CASA 10h observation with MeerKAT Measurement set
  36. 37 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT CASA 10h observation with MeerKAT Measurement set
  37. 38 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT
  38. 39 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT
  39. 40 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT Joint deconvolution ? Linear mosaicking ?
  40. 41 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT Joint deconvolution Linear mosaicking
  41. 42 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Validation

    using synthetic observations fBm with CASA and MeerKAT
  42. 44 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Mosaicking

    vs joint deconvolution Using ASKAP data Joint deconvolution Linear mosaicking
  43. 45 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Mosaicking

    vs joint deconvolution Using ASKAP data Joint deconvolution Linear mosaicking
  44. 46 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Mosaicking

    vs joint deconvolution Using ASKAP data Joint deconvolution Linear mosaicking
  45. 47 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Mosaicking

    vs joint deconvolution Using ASKAP data Joint deconvolution Linear mosaicking
  46. 49 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging GASKAP-HI

    Full Survey: Green & Scalable ? ! ! ANR grant (2 Postdocs and 1 Research Engineer ) Stay tuned for job offers
  47. 50 Antoine MARCHAL | IViS: Towards Green, Scalable Imaging Surveys

    SKA era A brief (& biased) history of HI surveys (in the local Universe) Sorted by end-date of observations Science verification of SKA-Mid