0.25
0.5
0.75
1.25
1.5
1.75
2
Approximation and Inference using Latent Variable Sparse Linear Models
Published on Feb 01, 20084416 Views
A variety of Bayesian methods have recently been introduced for performing approximate inference using linear models with sparse priors. We focus on four methods that capitalize on latent structure in
Related categories
Chapter list
Approximation and Inference Using Latent Variable Sparse Linear Models00:00
Overview00:12
Sparse Linear Model02:07
Sparse Prior: 2D Example03:14
Practical Issues03:38
Latent Variable Models of Sparse Priors04:17
1d Example of Convex-Type Representation05:48
Four Possibilities for Approximation06:18
Method I: w-MAP07:19
Method II: g-MAP08:11
Method III: Convex Bounding09:55
Method IV: Variational Bayes11:09
Unification12:53
Unification Cont.14:39
Choosing a Model15:22
Optimization Issues17:12
Example Applications18:39
Maximally Sparse Representations19:18
Example20:35
Using g-MAP to Find w021:02
Two Criteria for Choosing f(g) - 121:39
Two Criteria for Choosing f(g) - 222:25
Result22:51
Associated ‘Full’ Model25:27
Notes about SBL (and RVMs)26:47
Experimental Design28:14
Problem29:47
One Heuristic Solution30:53
Extensions31:43
Non-Negative Sparse Coding31:55
Non-Negative Sparse Coding Cont.32:19
Empirical Example - 133:04
Empirical Example - 233:34
Classification34:01
Covariance Component Estimation34:01
Final Thoughts34:02
Thank You35:14