Improve a Pareto frontier approximation.


The paretogreedy command is invoked either as:


The paretogreedy command tries to improve an existing Pareto frontier approximation. It uses the bi-objective multipoint likelihood criterion and the number of breakpoints criterion to assess and improve maps in the frontier. The command starts from all the maps stored into the heap. The 2-change neighborhood (reversing any subsection of the map) is visited for each map. Each time a non dominated map is found, it is stored into the heap, updating the frontier. The algorithm stops when all the maps have been visited. The Anytime ratio controls the tradeoff between search speed and solution quality. The total neighborhood size is equal to $\frac{NbMarkers (NbMarkers - 1)}{2}$. It is partially explored, depending on the Anytime ratio, but at least $2*NbMarkers$ neighbors are explored for each map in the frontier.

Arguments :


nothing. All the map explored by the algorithm are candidate for the heap.

See also:

Thomas Schiex 2009-10-27