Near-Linear Time Algorithm with Near-Logarithmic Regret Per Switch for Mixable/Exp-Concave Losses.
Kaan GökcesuHakan GökcesuPublished in: CoRR (2021)
Keyphrases
- worst case
- learning algorithm
- detection algorithm
- dynamic programming
- objective function
- k means
- experimental evaluation
- computational complexity
- preprocessing
- optimization algorithm
- theoretical analysis
- high accuracy
- np hard
- times faster
- improved algorithm
- matching algorithm
- minimum length
- simulated annealing
- computational cost
- search space
- linear programming
- online learning
- expectation maximization
- cost function
- shortest path
- support vector
- recognition algorithm
- genetic algorithm
- neural network