en
0.25
0.5
0.75
1.25
1.5
1.75
2
Sparse Algorithms are Not Stable: A No-free-lunch Theorem
Published on Jan 16, 20133873 Views
We consider two widely used notions in machine learning, namely: sparsity and stability. Both notions are deemed desirable, and are believed to lead to good generalization ability. We show that these
Related categories
Chapter list
Sparse Algorithms are not Stable: a no free lunch theorem00:00
The underlying problem00:22
Sparsity and Stability (1)02:15
Sparsity and Stability (2)02:25
Sparsity and Stability (3)02:33
Sparsity and Stability (4)02:43
Sparsity and Stability (5)03:39
Sparsity of Solution (1)04:25
Sparsity of Solution (2)05:09
Sparsity of Solution (3)05:46
Sparsity of Solution (4)06:43
Sparsity of Solution (5)06:57
This Talk: Can’t Have Both (1)07:21
This Talk: Can’t Have Both (2)07:42
Lasso is not stable (1)08:21
Lasso is not stable (2)08:23
Lasso is not stable (3)08:36
Lasso is not stable (4)08:39
Lasso is not stable (5)09:55
Lasso is not stable: proof (1)10:31
Lasso is not stable: proof (2)11:38
Lasso is not stable: proof (3)11:42
Lasso is not stable: proof (4)12:31
Lasso is not stable: proof (5)13:29
Lasso is not stable: proof (6)13:53
A more general statement/proof (1)14:43
A more general statement/proof (2)15:18
A more general statement/proof (3)15:19
A more general statement/proof (4)15:29
A more general statement/proof (5)15:40
Definitions for the proof (1)15:57
Definitions for the proof (2)16:04
Definitions for the proof (3)16:36
Definitions for the proof (5)17:03
Definitions for the proof (4)17:04
General proof (1)17:16
General proof (2)17:18
General proof (3)17:54
General proof (4)18:08
General proof (5)18:15
Some other thoughts (1)18:19
Some other thoughts (2)19:08