|
Tracking experts that learn by evolving past posteriors
Wouter Koolen and Tim Erven, van
Proceedins of the 18th Annual Belgian-Dutch Conference on Machine Learning
,
2009.
AbstractA problem posed by Freund is how to efficiently
track a small pool of experts out
of a much larger set. This problem was
solved when Bousquet andWarmuth introduced
their mixing past posteriors (MPP)
algorithm in 2001.
In Freund’s problem the experts would
normally be considered black boxes.
However, in this paper we re-examine
Freund’s problem in case the experts have
internal structure that enables them to
learn. In this case the problem has two
possible interpretations: should the experts
learn from all data or only from
the subsequence on which they are being
tracked? The MPP algorithm solves the
first case. We generalise MPP to address
the second option. Our results apply to any
expert structure that can be formalised using
(expert) hidden Markov models. Curiously
enough, for our interpretation there
are two natural reference schemes: freezing
and sleeping. For each scheme, we
provide an efficient prediction strategy [Edit]
|
