Timeline for Kullback-Leibler divergence WITHOUT information theory
Current License: CC BY-SA 3.0
8 events
when toggle format | what | by | license | comment | |
---|---|---|---|---|---|
Jan 5, 2017 at 15:02 | comment | added | whuber♦ | +1 Re the comment thread: The mind boggles at the thought that any concept that cannot be reduced to a "number of bits" is useless. | |
Jan 4, 2017 at 23:13 | comment | added | kjetil b halvorsen♦ | @Xi'an: There must be some connection between that number 1/2 and power of the corresponding likelihood ratio test? | |
Dec 20, 2016 at 5:35 | history | edited | Xi'an | CC BY-SA 3.0 |
added 356 characters in body
|
Dec 20, 2016 at 5:31 | comment | added | Xi'an | @PaulUszak: if I tell you that the Kullaback-Leibler distance between a N(0,1) and a N(1,1) distribution is 1/2, how does this help? | |
Dec 19, 2016 at 21:52 | comment | added | Paul Uszak | Well I mean seeing some actual numbers. Theory is cute but the world runs on numbers. There are no examples of KL divergence that use actual numbers, so I'm drawn to the conclusion that it is a theory with no possible application. The OP discussed the length of messages in bits and data compression. I was referring to any example that had a number of bits in it... | |
Dec 13, 2016 at 0:16 | comment | added | Paul Uszak | Any possibility of seeing a numerical example of this? | |
Nov 16, 2016 at 20:55 | history | edited | Xi'an | CC BY-SA 3.0 |
deleted 5 characters in body
|
Nov 16, 2016 at 13:07 | history | answered | Xi'an | CC BY-SA 3.0 |