
By leveraging the concentration of contextual bandit tasks around a low-dimensional affine subspace, we can learn this structure via online principal component analysis to reduce the expected regret over the encountered bandits.


coremsg

exploiting-low-dimensional-subspace-structure-for-efficient-meta-learning-in-contextual-bandits


Exploiting Low-Dimensional Subspace Structure for Efficient Meta-Learning in Contextual Bandits
