decision trees protocols and the fourier entropy
play

Decision trees, protocols, and the Fourier Entropy-Influence - PowerPoint PPT Presentation

Decision trees, protocols, and the Fourier Entropy-Influence Conjecture Andrew Wan (Simons Institute) John Wright (CMU) Chenggang Wu (Tsinghua) Fourier basics Given a Boolean function , Fourier


  1. Our results If is computable by a read-k DT , then . If is computable DT with expected depth d , and satisfies , then . also proven by [CKLS13] The FEI + conjecture of [OT13] composes. also proven by [OT13]

  2. Our technique • Want to show for certain Boolean functions .

  3. Our technique • Want to show for certain Boolean functions . • Previous papers have studied the expression

  4. Our technique • Want to show for certain Boolean functions . • Previous papers have studied the expression • We instead take an information theoretic approach via the Shannon Source Coding Theorem .

  5. Shannon Source Coding Theorem Given a random variable , avg # of bits needed to communicate .

  6. Shannon Source Coding Theorem Given a random variable , avg # of bits needed to communicate . Thus, to show , we need to construct an efficient protocol for communicating the value of .

  7. Shannon Source Coding Theorem Given a random variable , avg # of bits needed to communicate . Thus, to show , we need to construct an efficient protocol for communicating the value of . ( efficient = bits on average)

  8. a protocol for read-k DTs

  9. Protocol for read-k DTs is computed by 0 0 1 1 0 1 0 1 0 1

  10. Protocol for read-k DTs is computed by (a read-2 DT) 0 0 1 1 0 1 0 1 0 1

  11. Protocol for read-k DTs is computed by 0 0 1 1 0 1 0 1 0 1

  12. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1

  13. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  14. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  15. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  16. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  17. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  18. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  19. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  20. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  21. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  22. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  23. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  24. Protocol for read-k DTs is computed by Key fact : if is in the support 0 of , then the coordinates of appear in a root-to-leaf path in 0 1 1 . 0 1 0 1 0 1 Some sets in the support of :

  25. Protocol for read-k DTs Given , what should our protocol output? 0 0 1 1 0 1 0 1 0 1

  26. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 0 1 1 0 1 0 1 0 1

  27. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 0 1 0 1 0 1

  28. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 0 1 0 1 0 1

  29. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 2.) output the path’s description : 0 1 0 1 0 1

  30. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 2.) output the path’s description : 0 1 0 1 0 1

  31. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 2.) output the path’s description : 3.) indicate which nodes fall in : 0 1 0 1 0 1

  32. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 2.) output the path’s description : 3.) indicate which nodes fall in : 0 1 0 1 0 1

  33. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 2.) output the path’s description : 3.) indicate which nodes fall in : 0 1 0 1 0 1

  34. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 2.) output the path’s description : 3.) indicate which nodes fall in : 0 1 0 1 0 1 4.) final output:

  35. Protocol for read-k DTs Given , what should our protocol output? 0 0 1 1 0 1 0 1 0 1

  36. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 0 1 1 0 1 0 1 0 1

  37. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 0 1 0 1 0 1

  38. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 lots of choices! 0 1 0 1 0 1

  39. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 lots of choices! 0 1 0 1 0 1

  40. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 lots of choices! 0 1 0 1 0 1

  41. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find a path containing 0 1 1 lots of choices! 0 1 0 1 0 1

  42. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find the shortest path 0 1 1 containing 0 1 0 1 0 1

  43. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find the shortest path 0 1 1 containing 0 1 0 1 0 1

  44. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find the shortest path 0 1 1 containing 2.) output the path’s description : 0 1 0 1 0 1

  45. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find the shortest path 0 1 1 containing 2.) output the path’s description : 0 1 0 1 0 1 3.) indicate which nodes fall in :

  46. Protocol for read-k DTs Given , what should our protocol output? e.g., 0 1.) find the shortest path 0 1 1 containing 2.) output the path’s description : 0 1 0 1 0 1 3.) indicate which nodes fall in : 4.) final output:

  47. Analysis of protocol • A decision tree should be arranged with the most influential variables near the top.

  48. Analysis of protocol • A decision tree should be arranged with the most influential variables near the top. • Since every path output is root-to-leaf, the variables near the top will contribute a lot of bits to the expectation.

  49. Analysis of protocol • A decision tree should be arranged with the most influential variables near the top. • Since every path output is root-to-leaf, the variables near the top will contribute a lot of bits to the expectation. • In summary, contributes a lot to the expectation ⇒ is near the top of the tree ⇒ is highly influential

  50. A bad tree Let be a decision tree.

  51. A bad tree Let be a decision tree.

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend