Sequential complexities and uniform martingale LLN

Add to your list(s) Download to your calendar using vCal

• Alexander Rakhlin, University of Pennsylvania, The Wharton School
• Monday 23 June 2014, 11:30-12:00
• Centre for Mathematical Sciences, Meeting Room 2.

If you have a question about this talk, please contact clc32.

Uniform laws of large numbers play a key role in statistics and learning theory. In this talk, we describe martingale analogues of the uniform laws and introduce new notions of ``sequential complexities’’. Extending the symmetrization technique to sequences of dependent random variables leads us to a notion of a tree. We then introduce a definition of a tree covering number, extend the chaining analysis, introduce an analogue of the VC dimension, and prove a counterpart to the classical combinatorial result of Vapnik-Chervonenkis-Sauer-Shelah. Our definitions and results can be seen as non-i.i.d. extensions of some of the key notions in empirical process theory.

This talk is part of the Probability Theory and Statistics in High and Infinite Dimensions series.

This talk is included in these lists:

• All CMS events
• All Talks (aka the CURE list)
• CMS Events
• Centre for Mathematical Sciences, Meeting Room 2
• DPMMS Lists
• DPMMS info aggregator
• DPMMS lists
• Hanchen DaDaDash
• Interested Talks
• School of Physical Sciences
• Statistical Laboratory info aggregator
• bld31

Note that ex-directory lists are not shown.