Improved Confidence Bounds for the Linear Logistic Model
                and Applications to Bandits

                Kwang-Sung Jun 1 Lalit Jain 2 Blake Mason 3 Houssam Nassif 4

             Abstract               a feature vector for each arm. A common assumption is
                                that the reward is a noisy linear measurement of the under-
    We propose improved fixed-design confidence       lying feature vector of the arm being pulled. In other words,
    bounds for the lin ...