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.

Transcript

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

   View Slide

2. who am I?
   / /
   what’ve I been up to?
   1
   

   View Slide

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

   View Slide

4. implementation.
   

   View Slide

5. data
   = >
   physics
   

   View Slide

6. open source software for astrophysics
   2
   

   View Slide

7. why?
   

   View Slide

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

   View Slide

9. my open source contributions
   3
   

   View Slide

10. View Slide

11. gaussian processes
    4
    

    View Slide

12. p(data|physics)
    

    View Slide

13. data ~ N(model; noise)
    

    View Slide

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)
    

    View Slide

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)
    

    View Slide

16. data ~ N(model; noise)
    

    View Slide

17. data ~ N(model; noise)
    

    View Slide

18. so. why not?
    

    View Slide

19. data ~ N(model; noise)
    

    View Slide

20. View Slide

21. reference: Ambikasaran, DFM+ (2015)
    

    View Slide

22. View Slide

23. reference: Ambikasaran, DFM+ (2015)
    

    View Slide

24. reference: DFM, Agol, Ambikasaran, Angus (2017); DFM (2018); DFM, Luger, et al. (2021)
    

    View Slide

25. View Slide

26. reference: Gordon, Agol, DFM (2020)
    

    View Slide

27. what’s next?
    

    View Slide

28. View Slide

29. View Slide

30. View Slide

31. credit: Quang Tran
    

    View Slide

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

    View Slide

33. probabilistic inference
    5
    

    View Slide

34. p(data|physics)
    

    View Slide

35. have:
    
    
    physics
    = >
    data
    

    View Slide

36. want:
    
    
    data
    = >
    physics
    

    View Slide

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

    View Slide

38. View Slide

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

    View Slide

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

    View Slide

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

    View Slide

42. View Slide

43. View Slide

44. View Slide

45. View Slide

46. gradients!
    

    View Slide

47. dp(data|physics) / dphysics
    

    View Slide

48. automatic differentiation
    aka “backpropagation”
    

    View Slide

49. your model is just code
    

    View Slide

50. apply the chain rule
    

    View Slide

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

    View Slide

52. sounds silly?
    

    View Slide

53. it's not! (mostly)
    

    View Slide

54. View Slide

55. View Slide

56. what’s next?
    

    View Slide

57. View Slide

58. jax.readthedocs.io
    

    View Slide

59. my approach to open source
    6
    

    View Slide

60. View Slide

61. [1] don’t underestimate users
    
    
    [2] build libraries, not (just) scripts
    
    
    [3] teach by example
    

    View Slide

62. View Slide

63. View Slide

64. View Slide

65. bringing open source practices
    
    
    to research more generally
    

    View Slide

66. View Slide

67. View Slide

68. View Slide

69. View Slide

70. what’s next?
    7
    

    View Slide

71. 7
    [1] inference with stochastic or
    
    
    intractable models
    
    
    [2] what can we do to better support
    
    
    open source in astrophysics
    

    View Slide

72. View Slide

73. 7
    credit: Adrian Price-Whelan
    

    View Slide

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

    View Slide

75. a selection of some* CCA-supported software:
    * my apologies for neglecting your favorites!
    

    View Slide

76. View Slide

77. BUILDING
    
    
    THE
    
    
    SOFTWARE
    INFRASTRUCTURE
    
    
    FOR
    
    
    ASTROPHYSICS @ CCA
    

    View Slide