Measures of Statistical Dependence thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Measures of Statistical Dependence

Published on Feb 25, 200711041 Views

A number of important problems in signal processing depend on measures of statistical dependence. For instance, this dependence is minimised in the context of instantaneous ICA, in which linearly mixe

Related categories

Chapter list

ICA and Kernel Distribution Testing00:00
Overview00:52
Some notation and conventions02:10
ICA<br>... where to be careful when doing it02:54
ICA (Population version)03:00
ICA (empirical version)03:52
A toy example (1)04:04
A toy example (2)04:26
Things that are impossible for ICA (1)04:45
Things that are impossible for ICA (1) 0105:17
Things that are impossible for ICA (2)05:37
Things that are impossible for ICA (3)06:10
Things that are impossible for ICA (4)06:42
ICA Step 1 - <br>Decorrelation07:15
First step in ICA: decorrelate07:20
Example: what does decorrelation achieve?07:34
Decorrelation: a drawback08:03
What is left: rotation08:48
Rotation (continued)09:00
ICA: maximum likelihood09:19
Maximum likelihood: example10:48
Maximum likelihood: where it fails11:21
ICA Step 2(b) <br>Rotation: contrast functions12:21
What is a copy?12:30
Contrast functions13:03
Contrast functions and maximum likelihood14:08
Contrast functions and mutual information (1)15:05
Contrast functions and mutual information (2)15:44
Contrast functions (3): Some famous cases16:52
Kurtosis: an important concept17:32
Contrast functions: Example (1)18:20
Contrast functions: Example (2)18:40
Disclaimer!19:37
ICA for non-i.i.d. signals (1)20:14
ICA for non-i.i.d. signals (2)22:10
Advanced (kernel!) - <br>independence measures23:21
Kernel dependence measures23:25
Outline24:50
Dependence detection26:05
A second order method26:35
Take nonlinear features27:20
The kernel trick (1)29:30
The kernel trick (2)30:12
An empirical estimate30:46
COCO measures independence31:02
Why universal?31:33
Background: statistical tests (1)33:10
Background: statistical tests (2)35:28
When is dependence hard to detect?37:23
Hard-to-detect dependence (2)39:48
Hard-to-detect dependence (3)41:27
Hard-to-detect dependence (4)42:38
A test of independence43:12
Choosing kernel size (1)45:48
Choosing kernel size (2)47:51
Application to ICA49:02
Positive, Negative, and Zero kurtosis50:34
Outlier resistance51:46
The Two-Sample Problem52:25
The two-sample problem52:31
The MMD (1)53:41
The MMD (2)54:31
The MMD (2) 0156:05
Empirical estimate57:15