en
0.25
0.5
0.75
1.25
1.5
1.75
2
Efficient algorithms for estimating multi-view mixture models
Published on Jan 16, 20133008 Views
Mixture models are a staple in machine learning and applied statistics for treating data taken from multiple sub-populations. For many classes of mixture models, parameter estimation is computation
Chapter list
Efficient algorithms for estimating multi-view mixture models00:00
Outline00:02
Unsupervised learning00:22
Mixture models01:30
Multi-view mixture models - 102:33
Multi-view mixture models - 202:53
Multi-view mixture models - 303:36
Multi-view mixture models - 404:06
Semi-parametric estimation task - 104:56
Semi-parametric estimation task - 205:49
Some barriers to efficient estimation06:19
Making progress: discrete hidden Markov models10:53
What we do11:57
Part 2. Multi-view method-of-moments14:00
The plan - 114:39
The plan - 215:09
Simpler case: exchangeable views - 115:16
Simpler case: exchangeable views - 215:46
Key ideas17:12
Algebraic structure in moments18:17
2nd moment: subspace spanned by conditional means20:44
3rd moment: (cross) skew maximizers22:01
Variational analysis - 124:06
Variational analysis - 224:25
Variational analysis - 324:34
Variational analysis - 424:36
Variational analysis - 524:54
Variational analysis - 624:57
Variational analysis - 725:08
Variational analysis - 825:52
Extracting all isolated local maximizers26:25
General case: asymmetric views - 128:02
General case: asymmetric views - 228:14
Asymmetric cross moments28:35
Part 3. Some applications and open questions29:56
Mixtures of axis-aligned Gaussians30:05
Hidden Markov models and others31:18
Bag-of-words clustering model - 132:48
Bag-of-words clustering model - 233:33
Bag-of-words clustering model - 333:38
Bag-of-words clustering model - 433:40
Concluding remarks37:05
Thanks!38:20