Active and Adaptive Sequential Learning with Per Time-step Excess Risk Guarantees | IEEE Conference Publication | IEEE Xplore