Performance Metrics. We report the accuracy with which the
group of a new user is estimated i.e. the fraction of times the cor-
rect group is estimated and also the regret i.e. ∑𝑛 𝑖=1 𝑟 (𝑣𝑖 ) − 𝑟 (𝑣∗𝑖 )
where 𝑣∗
𝑖 is the unrated item with highest rating by the new user
(so 𝑣∗
1 is the highest rated item, 𝑣∗
2 the next highest and so on). We
measure the convergence time as the number of items rated before
the regret reaches 80% of its final value over a test run. Statistics
are calculated over 1000 new users per group.
Performance Metrics. We report the accuracy with which the group of a new user is estimated i.e. the fraction of times the cor- rect group is estimated and also the regret i.e. ∑𝑛 𝑖=1 𝑟 (𝑣𝑖 ) − 𝑟 (𝑣∗𝑖 ) where 𝑣∗ 𝑖 is the unrated item with highest rating by the new user (so 𝑣∗ 1 is the highest rated item, 𝑣∗ 2 the next highest and so on). We measure the convergence time as the number of items rated before the regret reaches 80% of its final value over a test run. Statistics are calculated over 1000 new users per group.