Learning Good Edit Similarities with Generalization Guarantees thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Learning Good Edit Similarities with Generalization Guarantees

Published on Nov 30, 20112716 Views

Similarity and distance functions are essential to many learning algorithms, thus training them has attracted a lot of interest. When it comes to dealing with structured data (e.g., strings or trees),

Related categories

Chapter list

Learning Good Edit Similarities with Generalization Guarantees00:00
Introduction: Similarity Learning00:52
Similarity functions in classification00:55
Similarity learning01:23
Goals of our work02:13
(ǫ, γ, τ ) - good similarity functions02:47
Definition - 102:48
Definition - 203:22
Intuition behind the definition - 104:45
Intuition behind the definition - 205:46
Implications for learning - 106:02
Learning rule07:02
L1 - norm and Sparsity07:38
Learning good edit similarities08:29
Motivations for our work08:43
The string edit distance09:29
A feel of the state-of-the-art in edit cost learning - 110:44
A feel of the state-of-the-art in edit cost learning - 211:00
A feel of the state-of-the-art in edit cost learning - 311:19
A feel of the state-of-the-art in edit cost learning - 411:34
Our edit similarity function - 112:09
Our edit similarity function - 212:35
Our edit similarity function - 313:03
Optimize the goodness13:11
Optimize the goodness ctd14:11
Convex formulation of the problem15:01
Learning guarantees16:04
Convergence and learning guarantees16:53
Convergence rate: accuracy17:31
Classification performance: sparsity18:42
Conclusions19:04