SuperLectures.com

DISTRIBUTED TRAINING OF LARGE SCALE EXPONENTIAL LANGUAGE MODELS

Language Modeling

Full Paper at IEEE Xplore

Presented by: Bhuvana Ramabhadran, Author(s): Abhinav Sethy, Stanley Chen, Bhuvana Ramabhadran, IBM, United States

Shrinkage-based exponential language models, such as the recently introduced Model M, have provided significant gains over a range of tasks . Training such models requires a large amount of computational resources in terms of both time and memory. In this paper, we present a distributed training algorithm for such models based on the idea of cluster expansion . Cluster expansion allows us to efficiently calculate the normalization and expectations terms required for Model M training by minimizing the computation needed between consecutive n-grams. We also show how the algorithm can be implemented in a distributed environment, greatly reducing the memory required per process and training time.

  Speech Transcript

|

  Slides

Enlarge the slide | Show all slides in a pop-up window

0:00:16

  1. slide

0:00:35

  2. slide

0:01:07

  3. slide

0:01:58

  4. slide

0:02:31

  5. slide

0:03:15

  6. slide

0:03:44

  7. slide

0:04:53

  8. slide

0:06:01

  9. slide

0:06:22

 10. slide

0:07:27

 11. slide

0:08:28

 12. slide

0:09:17

 13. slide

0:09:49

 14. slide

0:10:10

 15. slide

0:11:32

 16. slide

0:12:35

 17. slide

0:13:08

 18. slide

0:13:38

 19. slide

0:14:09

 20. slide

0:14:20

 21. slide

0:14:28

 22. slide

0:15:05

    20. slide

0:15:25

 23. slide

0:17:07

 24. slide

  Comments

Please sign in to post your comment!

  Lecture Information

Recorded: 2011-05-25 16:35 - 16:55, Club H
Added: 9. 6. 2011 01:58
Number of views: 47
Video resolution: 1024x576 px, 512x288 px
Video length: 0:19:16
Audio track: MP3 [6.58 MB], 0:19:16