Sciweavers

COLT
2008
Springer
15 years 3 days ago
More Efficient Internal-Regret-Minimizing Algorithms
Standard no-internal-regret (NIR) algorithms compute a fixed point of a matrix, and hence typically require O(n3 ) run time per round of learning, where n is the dimensionality of...
Amy R. Greenwald, Zheng Li, Warren Schudy