On 16.04.2015 12:16, haitam wrote:
Hi All, I would like to know whether with the actual A-star, we can be able to develop a multi-objective approach. After a brief introduction I came to know that it is possible to have a vector weight, and customize the different steps of the algorithm with a Visitor. Now I am wondering if: + the objective function (distance) can only be a scalar and not a vector ? + Is the algorithm (behind the scenes) bear a multi-objective implementation, for instance for A* a tree is constructed, but for MOA* rather an Acyclic Graph is constructed.
Multiobjective A* with vector weights is not implemented.
If not supported, I should maybe be add a new search algorithm to the set. What is the best way to extend Graph Tool ? Do I have only to create the graph_moastar.cc, graph_moastar.hh, graph_moastar.lo and update the Makefile in the src/graph/search folder ?
Yes, as well as the python code in src/graph_tool/search/__init__.py. Just take at look at the astar code, and reproduce its structure. Best, tiago -- Tiago de Paula Peixoto <tiago@skewed.de>