Theory-inspired parameter control benchmarks for dynamic algorithm configuration.
André BiedenkappNguyen DangMartin S. KrejcaFrank HutterCarola DoerrPublished in: GECCO (2022)
Keyphrases
- learning algorithm
- improved algorithm
- dynamic programming
- theoretical analysis
- parameter settings
- recognition algorithm
- times faster
- detection algorithm
- high accuracy
- particle swarm optimization
- estimation algorithm
- path planning
- optimization algorithm
- computationally efficient
- objective function
- optimal solution
- preprocessing
- computational cost
- linear programming
- k means
- hidden markov models
- benchmark suite
- control parameters
- parameter values
- cost function
- worst case
- segmentation algorithm
- markov random field
- simulated annealing
- graph cuts
- genetic algorithm