en
0.25
0.5
0.75
1.25
1.5
1.75
2
Sparsity regret bounds for individual sequences in online linear regression
Published on 2011-08-023374 Views
We consider the problem of online linear regression on arbitrary deterministic sequences when the ambient dimension d can be much larger than the number of time rounds T. We introduce the notion of sp
Related categories
Presentation
Sparsity regret bounds for individual sequences in online linear regression00:00
Introduction: sparsity in the stochastic setting00:08
Outline02:25
Introduction of the notion of sparsity regret bound03:02
Setting: online linear regression on individual sequences03:06
Prediction goal04:32
What about the high dimensions?05:47
Sparsity regret bounds06:43
Related works in online convex optimization08:02
Main results with individual sequences09:39
Algorithm: Sequential Sparse Exponential Weighting (SeqSEW)10:02
Known bound By on the observations12:02
Unknown bound By on the observations13:29
Adaptivity results with i.i.d. data15:06
Application to i.i.d. data15:15
Adaptivity to the unknown variance - 116:32
Adaptivity to the unknown variance - 217:06
Conclusion and ongoing work - 118:25
Conclusion and ongoing work - 219:14