Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Tunable algorithms for transient follow-up

279 views

Published on

In this talk I introduce Bayesian decision theory, in the context of transient follow-up prioritisation and automated scheduling with robotic astronomical observatories.

Published in: Science
  • Be the first to comment

  • Be the first to like this

Tunable algorithms for transient follow-up

  1. 1. Tunable algorithms for transient follow-up Tim Staley TKP Meeting, Manchester, Sept 2014 WWW: 4pisky.org , timstaley.co.uk
  2. 2. Context Theory Implementation Future work Fin Aim of this talk A basic, intuitive understanding of information content and how this can be used to optimize / automate decision making, a.k.a. Bayesian decision theory
  3. 3. Context Theory Implementation Future work Fin Outline Context Theory Implementation Future work Fin
  4. 4. Context Theory Implementation Future work Fin A blueprint for automated follow-up
  5. 5. Context Theory Implementation Future work Fin Outline Context Theory Implementation Future work Fin
  6. 6. Context Theory Implementation Future work Fin 6 4 2 0 2 4 6 Epoch 0.0 0.2 0.4 0.6 0.8 1.0 Relativeflux stable logistic null Intrinsic lightcurves
  7. 7. Context Theory Implementation Future work Fin 6 4 2 0 2 4 6 Epoch 0.5 0.0 0.5 1.0 1.5 Relativeflux True value Noisy samples Sampling with noise
  8. 8. Context Theory Implementation Future work Fin 6 4 2 0 2 4 6 Epoch 0.5 0.0 0.5 1.0 1.5 Relativeflux Sampling with noise
  9. 9. Context Theory Implementation Future work Fin 6 4 2 0 2 4 6 Epoch 0.5 0.0 0.5 1.0 1.5 Relativeflux Sampling with noise
  10. 10. Context Theory Implementation Future work Fin 0 1 0.5 0.0 0.5 1.0 1.5Relativeflux T=-5.0 0 1 T=-4.0 0 1 PDF value T=-3.0 0 1 T=-2.0 0 1 T=-1.0 0 1 T=0.0 0 1 T=1.0 0 1 T=2.0 stable logistic null Class PDF at each epoch
  11. 11. Context Theory Implementation Future work Fin PDF value0.5 0.0 0.5 1.0 1.5 Relativeflux T=-5.0 T=-4.0 T=-3.0 T=-2.0 T=-1.0 T=0.0 T=1.0 T=2.0 stable logistic null 5 4 3 2 1 0 1 2 Epoch 0.50 0.45 0.40 0.35 0.30 0.25 FoM Information content Evaluating each epoch
  12. 12. Context Theory Implementation Future work Fin Confusion matrices
  13. 13. Context Theory Implementation Future work Fin Confusion matrices
  14. 14. Context Theory Implementation Future work Fin −4 −2 0 2 4 Time 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Featurevalue stable logistic Instrinsic lightcurves - ensemble
  15. 15. Context Theory Implementation Future work Fin Outline Context Theory Implementation Future work Fin
  16. 16. Context Theory Implementation Future work Fin Required knowledge / user-inputs Transient rate priors.
  17. 17. Context Theory Implementation Future work Fin Required knowledge / user-inputs Transient rate priors. Transient lightcurve ensemble models.
  18. 18. Context Theory Implementation Future work Fin Required knowledge / user-inputs Transient rate priors. Transient lightcurve ensemble models. Telescope / noise models.
  19. 19. Context Theory Implementation Future work Fin Required knowledge / user-inputs Transient rate priors. Transient lightcurve ensemble models. Telescope / noise models. Follow-up prioritization weightings.
  20. 20. Context Theory Implementation Future work Fin Required software components Efficient lightcurve generation library.
  21. 21. Context Theory Implementation Future work Fin Required software components Efficient lightcurve generation library. MCMC data fitting models and routines.
  22. 22. Context Theory Implementation Future work Fin Required software components Efficient lightcurve generation library. MCMC data fitting models and routines. Statistical routines for calculating confusion matrices.
  23. 23. Context Theory Implementation Future work Fin Required software components Efficient lightcurve generation library. MCMC data fitting models and routines. Statistical routines for calculating confusion matrices. Observation schedule optimization engine.
  24. 24. Context Theory Implementation Future work Fin Required components
  25. 25. Context Theory Implementation Future work Fin Outline Context Theory Implementation Future work Fin
  26. 26. Context Theory Implementation Future work Fin What’s next? Finish bolting components together. Run simulations, test in more realistic scenarios. Interfacing with optimizer / scheduler.
  27. 27. Context Theory Implementation Future work Fin Longer term Variational Bayes? Gaussian processes?
  28. 28. Context Theory Implementation Future work Fin Outline Context Theory Implementation Future work Fin
  29. 29. Context Theory Implementation Future work Fin Summary Information content is just a penalty function for scoring predicted observations. Using it to decide when to observe is applied Bayesian decision theory. But doing this for real requires a number of non-trivial software components. Nearly ready for testing!

×