It was unexpectedly (this is maybe not the correct word to use, beyond my expectation might be more appropriate?) successful of the multi-armed bandit workshop mainly organized by Steffen Grünewälder, hosted at Imperial College London. It is rather rare (to the best of my knowledge, it is unprecedented) to have a pure workshop dedicated to such a small sub-field like ours. The workshop has gathered, I would say, almost half of the known researchers in the domain, mostly based on Europe, but some of whom even come from north America. The two-day workshop is composed of several invited talks and one poster session each day.
It was a great pleasure for me to meet a bunch of brilliant young new comers on the topic, but also to re-unite with some old SequeL (my team) friends. It is simultaneously surprising and unsurprising to see that almost 1/3 of the present attendees have or had some links to SequeL. There are two papers that seem to interest me a lot: “Non-Asymptotic Pure Exploration by Solving Games” by Rémy Degenne, Wouter M. Koolen and Pierre Ménard, and “Best Arm Identification in Spectral Bandits” by Tomáš Kocák and Aurélien Garivier
(it should be Best-Arm instead of Best Arm in the title). I am recently starting to work on pure exploration for linear bandits, which should be highly correlated to the mentioned two papers. In particular, it would be interesting to investigate the game-theoretical point of view in the BAI for linear bandits case.
Fun fact: The entry of the Maths building of Imperial College is outside the campus.