An Improved Parametrization and Analysis of the EXP3++ Algorithm for Stochastic and Adversarial Bandits.
Yevgeny SeldinGábor LugosiPublished in: CoRR (2017)
Keyphrases
- experimental evaluation
- dynamic programming
- learning algorithm
- times faster
- computational complexity
- optimal solution
- detection algorithm
- np hard
- computationally efficient
- high accuracy
- linear programming
- convex hull
- optimization algorithm
- theoretical analysis
- expectation maximization
- worst case
- cost function
- multi objective
- significant improvement
- artificial neural networks
- input data
- monte carlo
- matching algorithm
- objective function