$30 off During Our Annual Pro Sale. View Details »

My research talk for CCA promotion

My research talk for CCA promotion

A summary of what I've been up to for the past few years and where my research program is going.

Dan Foreman-Mackey

February 03, 2022
Tweet

More Decks by Dan Foreman-Mackey

Other Decks in Science

Transcript

  1. BUILDING THE SOFTWARE INFRASTRUCTURE FOR ASTROPHYSICS by Dan Foreman-Mackey

  2. who am I? / / what’ve I been up to?

    1
  3. 7 [1] solving Hard™ data analysis problems [2] enabling and

    empowering astrophysicists
  4. implementation.

  5. data = > physics

  6. open source software for astrophysics 2

  7. why?

  8. credit: Adrian Price-Whelan / / data: SAO/NASA ADS

  9. my open source contributions 3

  10. None
  11. gaussian processes 4

  12. p(data|physics)

  13. data ~ N(model; noise)

  14. °0.6 °0.3 0.0 0.3 0.6 raw [ppt] 0 5 10

    15 20 25 time [days] °0.30 °0.15 0.00 de-trended [ppt] N = 1000 reference: DFM+ (2017)
  15. °0.6 °0.3 0.0 0.3 0.6 raw [ppt] 0 5 10

    15 20 25 time [days] °0.30 °0.15 0.00 de-trended [ppt] N = 1000 reference: DFM+ (2017)
  16. data ~ N(model; noise)

  17. data ~ N(model; noise)

  18. so. why not?

  19. data ~ N(model; noise)

  20. None
  21. reference: Ambikasaran, DFM+ (2015)

  22. None
  23. reference: Ambikasaran, DFM+ (2015)

  24. reference: DFM, Agol, Ambikasaran, Angus (2017); DFM (2018); DFM, Luger,

    et al. (2021)
  25. None
  26. reference: Gordon, Agol, DFM (2020)

  27. what’s next?

  28. None
  29. None
  30. None
  31. credit: Quang Tran

  32. reference: Luger, DFM, Hedges (2021)

  33. probabilistic inference 5

  34. p(data|physics)

  35. have: physics = > data

  36. want: data = > physics

  37. integral of the form f(physics) p(physics|data) dphysics

  38. None
  39. number of parameters patience required a few tenish not outrageously

    many reference: DFM (priv. comm.)
  40. number of parameters patience required emcee a few tenish not

    outrageously many reference: DFM (priv. comm.)
  41. number of parameters patience required emcee a few tenish not

    outrageously many how things should be reference: DFM (priv. comm.)
  42. None
  43. None
  44. None
  45. None
  46. gradients!

  47. dp(data|physics) / dphysics

  48. automatic differentiation aka “backpropagation”

  49. your model is just code

  50. apply the chain rule

  51. apply the chain rule over and over again . .

    .
  52. sounds silly?

  53. it's not! (mostly)

  54. None
  55. None
  56. what’s next?

  57. None
  58. jax.readthedocs.io

  59. my approach to open source 6

  60. None
  61. [1] don’t underestimate users [2] build libraries, not (just) scripts

    [3] teach by example
  62. None
  63. None
  64. None
  65. bringing open source practices to research more generally

  66. None
  67. None
  68. None
  69. None
  70. what’s next? 7

  71. 7 [1] inference with stochastic or intractable models [2] what

    can we do to better support open source in astrophysics
  72. 7

  73. 7 credit: Adrian Price-Whelan

  74. many fundamental software packages have a shockingly small number of

    maintainers.
  75. a selection of some* CCA-supported software: * my apologies for

    neglecting your favorites!
  76. None
  77. BUILDING THE SOFTWARE INFRASTRUCTURE FOR ASTROPHYSICS @ CCA