1 / 61

Imputing Missing Events in Continuous-Time Event Streams

Explore the challenges and solutions of imputing missing events in dynamic event streams through sequential Monte Carlo methods. Learn about neural Hawkes processes and inference strategies for solving incomplete data scenarios.

amandas
Download Presentation

Imputing Missing Events in Continuous-Time Event Streams

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Imputing Missing Events in Continuous-Time Event Streams Guanghui Qin2 Jason Eisner1 Hongyuan Mei1 1Department of Computer Science, Johns Hopkins University, USA 2Department of Physics, Peking University, China

  2. Event Streams Medical

  3. Event Streams Shopping Medical

  4. Event Streams Social media Shopping Medical

  5. Event Streams Social media Shopping Medical

  6. Event Streams Social media Shopping Medical

  7. Let’s Play a Battle Game

  8. Let’s Play a Battle Game • Incomplete data

  9. Let’s Play a Battle Game • Incomplete data • Opponent tanks come

  10. Let’s Play a Battle Game • Incomplete data • Opponent tanks come

  11. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them?

  12. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built?

  13. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? ?

  14. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? ?

  15. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? ? ?

  16. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? • Where was it built? ? ?

  17. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? • Where was it built? ? ?

  18. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? • Where was it built? ? ? ?

  19. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? • Where was it built? ? ? ?

  20. Let’s Play a Battle Game • Incomplete data • Opponent tanks come • Maybe a new factory was built to produce them? • When was it built? • Where was it built? ? ? ?

  21. Why to Impute? ? ? ? ? ?

  22. Why to Impute? • If there is a factory over there • We should go after it!

  23. To Impute the Missing Events…

  24. To Impute the Missing Events… • We need to know the distribution of the complete sequences

  25. To Impute the Missing Events… • We need to know the distribution of the complete sequences Neural Hawkes process (NHP) (Mei & Eisner 2017 – NeurIPS)

  26. To Impute the Missing Events… • Where do we get the complete data to train its parameters?

  27. To Impute the Missing Events… • Where do we get the complete data to train its parameters? • Game logs shown to us after the game (e.g. replay)

  28. To Impute the Missing Events… • We also need to know the missingness mechanism • Don’t propose anything which we know won’t be missing

  29. To Impute the Missing Events… • We also need to know the missingness mechanism • Don’t propose anything which we know won’t be missing • In-view events won’t be missing

  30. To Impute the Missing Events… • We also need to know the missingness mechanism • Don’t propose anything which we know won’t be missing • In-view events won’t be missing • Out-of-view events must be missing

  31. To Impute the Missing Events…

  32. To Impute the Missing Events…

  33. To Impute the Missing Events…

  34. To Impute the Missing Events… ? ? ? ?

  35. To Impute the Missing Events… ? ? ? ?

  36. Challenge • Complicated • Exact inference is intractable • How about Monte Carlo? ? ? ? ?

  37. Sequential Monte Carlo • Sample from

  38. Sequential Monte Carlo • Sample from

  39. Sequential Monte Carlo • Sample from • Weight them by

  40. Sequential Monte Carlo • Sample from • Weight them by Rare to have two factories in a row!

  41. Sequential Monte Carlo • Sample from • Weight them by Down-weighted! Too late! No time to produce tanks!

  42. Sequential Monte Carlo • Sample from • Weight them by Upweighted!

  43. Sequential Monte Carlo • Use the trained NHP Infinitesimal probability of (like a Poisson intensity)

  44. Sequential Monte Carlo • Use the trained NHP

  45. Sequential Monte Carlo • Use the trained NHP Produce tanks so fast?!

  46. Sequential Monte Carlo • Use the trained NHP Low! No factory built!

  47. Sequential Monte Carlo • Use the trained NHP What produced the tanks?!

  48. Sequential Monte Carlo • Use the trained NHP • Take future into account Consider future for sampling!

  49. Sequential Monte Carlo • Use the trained NHP • Take future into account Read future with another LSTM

  50. Sequential Monte Carlo • Use the trained NHP • Take future into account Improved proposaldistribution

More Related