Drawbacks of Iterative Improvement Algorithms
NTT Communication Science Labs.
The theoretical completeness can not be guaranteed.
- The fill algorithm can always find a solution, but will not terminate when the problem is over-constrained.
- The breakout algorithm may fall into an infinite processing loop.
The applicability of existing heuristics is limited.
- Many powerful heuristics (e.g., forward checking) presuppose the existence of consistent partial solutions.