Paper detail

Chasing Ghosts: Competing with Stateful Policies

We consider sequential decision making in a setting where regret is measured with respect to a set of stateful reference policies, and feedback is limited to observing the rewards of the actions performed (the so called "bandit" setting). If either the reference policies are stateless rather than stateful, or the feedback includes the rewards of all actions (the so called "expert" setting), previous work shows that the optimal regret grows like $Θ(\sqrt{T})$ in terms of the number of decision rounds $T$. The difficulty in our setting is that the decision maker unavoidably loses track of the internal states of the reference policies, and thus cannot reliably attribute rewards observed in a certain round to any of the reference policies. In fact, in this setting it is impossible for the algorithm to estimate which policy gives the highest (or even approximately highest) total reward. Nevertheless, we design an algorithm that achieves expected regret that is sublinear in $T$, of the form $O( T/\log^{1/4}{T})$. Our algorithm is based on a certain local repetition lemma that may be of independent interest. We also show that no algorithm can guarantee expected regret better than $O( T/\log^{3/2} T)$.

preprint2014arXivOpen access

Signal facts

What is known right now

Open access3 authors1 topic

Next steps

Decide what to do with this paper

Use like or dislike for the fast social read. The more specific scholarly feedback stays available below when needed.

Log in to curate

Reading frame

Keep the important context close to the paper

Keep the important signals around this paper in one place: votes, save state, collection context, reviews and the metadata you need before deciding what to do next.

Institutions

Add specific reaction

Move through the context

Research map

Open full explorer

Move through nearby people, institutions, topics and adjacent work without leaving the paper page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Structured reviews

0 review(s)

ContributeLeave structured feedbackUse the review template when you have a concrete strength, concern or method question.Open review form

No structured reviews yet. High-signal critique starts here.

Work discussion

0 comment(s)

DiscussAdd a high-signal commentKeep quick notes, caveats and replication pointers separate from formal reviews.Open comment form

No discussion yet. The first strong comment sets the tone.