Search papers, labs, and topics across Lattice.
2
0
3
Surrogate probability functions enable FTPL to achieve BOBW guarantees while maintaining computational simplicity, revolutionizing adaptive learning rates in online learning.
FTPL finally achieves "Best-of-Both-Worlds" optimality for $m$-set semi-bandit problems, matching FTRL's regret guarantees with significantly reduced computational complexity via conditional geometric resampling.
