Can be viewed as a more sophisticated selection rule or a revised
set of local moves.
- Optimization performed in ``sweeps''.
- At the beginning of the sweep, mark all nodes as available.
- From all available nodes, move the one the improves the score the
most.
- Mark it as unavailable.
- Repeated until no nodes available.
- Begin the next sweep from best answer found on this sweep.
- Repeat sweeps until no improvement.
Really hard to fool it! Pushes through standard local optima.
Final Project: Apply this idea to SAT?
Next: SIMUALTED ANNEALING
Up: LOCAL SEARCH
Previous: Graph Partitioning