0.25
0.5
0.75
1.25
1.5
1.75
2
Universal Modeling: Introduction to modern MDL
Published on Feb 25, 200710609 Views
We give a tutorial introduction to the *modern* Minimum Description Length (MDL) Principle, taking into account the many refinements and developments that have taken place in the 1990s. These do not s
Related categories
Chapter list
Universal Modeling: Introduction to ‘Modern’ MDL00:00
Minimum Description Length Principle01:28
Minimum Description Length Principle02:04
Minimum Description Length Principle05:51
Model Selection06:19
Example: Regression07:35
Example: Regression07:54
Example: Regression08:23
Example: Regression08:35
Example: Regression08:44
Modern MDL!08:46
Five MDL Lectures09:40
Part I: Overview10:23
Codes10:56
Example 1: uniform code12:13
Code Length & Probability12:54
Code Lengths ‘are’ probabilities…14:07
…and probabilities ‘are’ code lengths!15:04
The Most Important Slide!15:40
The Most Important Slide!16:14
Example 1: uniform code/distr.17:09
Prefix codes distributions18:05
Prefix codes distributions19:41
General Recipe (Kraft)20:59
Prefix codes distributions21:39
Prefix codes distributions21:46
General Recipe (Kraft)22:07
Example 3: distributions codes23:16
Example 3: distributions codes24:38
The Most Important Slide!25:57
Remarks26:37
Part I: Overview27:39
Universal Codes27:52
Universal Codes29:11
Universal Codes30:51
Universal Codes33:04
Universal Codes33:15
Terminology37:06
Bayesian Mixtures are universal models38:01
Bayesian Mixtures are universal models38:55
2-part MDL code is a universal model (code)39:55
Bayesian Mixtures are universal models40:49
2-part MDL code is a universal model (code)41:36
2-part vs. Bayes universal models42:03
Optimal Universal Model42:06
Optimal Universal Model - II43:27
MDL Model Selection45:04
MDL Model Selection45:40