video thumbnail
Pause
Mute
Subtitles not available
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Learning Overcomplete Latent Variable Models through Tensor Methods

Published on Feb 4, 20253753 Views

We provide guarantees for learning latent variable models emphasizing on the overcomplete regime, where the dimensionality of the latent space exceeds the observed dimensionality. In particular, we co

Related categories

Presentation

Learning Overcomplete Latent Variable Models through Tensor Methods00:00
Latent Variable Probabilistic Models00:06
Method-of-Moments (Spectral methods)00:20
Multiview Mixture Model00:34
Tensor Rank and Tensor Decomposition01:31
Our Setup02:01
Alternating minimization02:45
Main Result: Local Convergence03:08
Global Convergence k = O(d)04:03