Iterative Improvement Algorithms
NTT Communication Science Labs.
No consistent partial solution is constructed.
A whole flawed solution is revised by hill-climbing search.
- Min-conflict heuristic can be used for guiding the search.
- Local-minima can be escaped by changing the weight of constraints (breakout), or by changing the evaluation values of the local-minima (fill) [Morris, 93]
A mistake can be revised without conducting an exhaustive search.