Graph Feedback via Reduction to Regression
share
When feedback is partial, leveraging all available information is critical to minimizing data requirements. Graph feedback, which interpolates between the supervised and bandit regimes, has been extensively studied; but the mature theory is grounded in impractical algorithms. We present and analyze an approach to contextual bandits with graph feedback based upon reduction to regression. The resulting algorithms are practical and achieve known minimax rates.
READ FULL TEXT