OSOM: A simultaneously optimal algorithm for multi-armed and linear contextual bandits.
Niladri S. ChatterjiVidya MuthukumarPeter L. BartlettPublished in: AISTATS (2020)
Keyphrases
- dynamic programming
- optimal solution
- detection algorithm
- linear complexity
- preprocessing
- closed form
- worst case
- learning algorithm
- high accuracy
- locally optimal
- times faster
- piecewise linear
- optimization algorithm
- optimal parameters
- linear programming
- computationally efficient
- cost function
- globally optimal
- significant improvement
- computational complexity
- convergence rate
- optimal strategy
- expectation maximization
- clustering method
- shortest path
- input data
- experimental evaluation
- np hard
- reinforcement learning