Upper-Confidence-Bound Action Selection

Action-Value Methods

$$A_t=\underset{a}{\text{arg max}}\left[Q_t(a)+c\sqrt{\frac{\ln t}{N_t(a)}}\right]$$