0.25
0.5
0.75
1.25
1.5
1.75
2
On the Usefulness of Similarity based Projection Spaces for Transfer Learning
Published on Oct 17, 20116401 Views
Similarity functions are widely used in many machine learning or pattern recognition tasks. We consider here a recent framework for binary classifi cation, proposed by Balcan et al., allowing to lea
Related categories
Chapter list
On the Usefulness of Similarity based Projection Spaces for Transfer Learning00:00
Introduction and Motivation00:00
Outline00:53
Outline: A Transfer Learning Task01:25
Formalisation01:28
Studied case - 102:16
Studied case - 202:54
Studied case - 302:59
S. Ben-David et al. results - 103:14
S. Ben-David et al. results - 203:29
S. Ben-David et al. results - 303:39
S. Ben-David et al. results - 404:08
S. Ben-David et al. results - 504:36
Outline: Learning with Good Similarity Functions (SF)04:53
Good Similarity Functions04:53
Properties05:41
Outline:Modifying the Projection Space for Domain Adaptation06:31
Modifying the Projection Space for Domain Adaptation: An Heuristic Normalization06:31
An Heuristic Normalization of a Similarity Function06:38
Modifying the Projection Space for Domain Adaptation: An Additional Regularization Term07:23
An Additional Regularization Term For Moving Closer the Two Distributions - 107:33
An Additional Regularization Term For Moving Closer the Two Distributions - 207:43
An Additional Regularization Term For Moving Closer the Two Distributions - 308:21
An Additional Regularization Term For Moving Closer the Two Distributions - 409:05
An Additional Regularization Term For Moving Closer the Two Distributions - 509:28
An Additional Regularization Term For Moving Closer the Two Distributions - 609:37
An Additional Regularization Term For Moving Closer the Two Distributions - 709:56
Reverse Classi er hr and Validation - 110:08
Reverse Classi er hr and Validation - 210:26
Reverse Classi er hr and Validation - 310:36
Reverse Classi er hr and Validation - 410:41
Reverse Classi er hr and Validation - 510:45
Reverse Classi er hr and Validation - 610:54
Reverse Classi er hr and Validation - 711:00
Outline: Experimentations11:13
Experimental Setup11:18
Inter-twinning moons: results12:22
Inter-twinning moons: estimation of the similarity function goodness on TS12:56
Images corpus: results14:05
Outline: Extended Work: A Little Bit of Theory14:25
Sparsity Analysis14:37
Generalization Bounds15:29
Outline: Conclusion and Perspectives16:16
Conclusion and Perspectives16:18
Thank you for your attention.17:13