Super‐heuristics and Their Applications to Combinatorial Problems

Abstract

Combinatorial problems are known to be difficult because of the shear size of the solution space and the lack of polynomial time algorithms to “solve” them. Heuristics are often devised to produce acceptable solutions in an affordable time. In this paper, we propose a method called super‐heuristic that expands the capabilities of heuristics using randomization and sampling techniques. We submit that heuristics are in general strategies that map from available information of a problem instance to decisions in solution constructions/improvement. We show that it is important to utilize the information effectively in the randomization process. More importantly, the possibility of randomization around a heuristic spells out the demarcation between the roles of human and machines in complex optimization problems.

Approximate Nonmyopic Sensor Selection via Submodularity and Partitioning

“…It includes an aggregative phase to heuristically choose the initial subset and an adaptative phase to iteratively aggregate and disaggregate the current subset until it converges. Kalandros et al [23] explore the use of randomization and superheuristics search [31] for sensor selections in target-tracking applications. The search begins with a base sensor set and then generates more alternative solutions via a probabilistic assignment rule.…”

Random Approximated Greedy Search for Feature Subset Selection

“…Considerable efforts have been made to improve its results. Greedy Randomized Adaptive Search Procedures (GRASP) [8] and Super-heuristics [9] methods propose similar ideas for this target. By introducing the randomness in the greedy search procedure, one can expect to escape from local optimum, and have a chance to obtain better result.…”