This function is based on Yen's k-Shortest Path algorithm:
J. Y. Yen, "Finding the K shortest loopless paths in a network", Management Science 17:712–716, 1971.
1) [shortestPaths]: the list of K shortest paths (in cell array 1xK)
2) [totalCosts] : costs of the K shortest paths (in array 1xK)
Yen's algorithm prevents loops.
This function calls a slightly modified/simplified function dijkstra() (submitted by Xiaodong Wang, 2004)
The Network/Graph of N nodes is fed in the form of a NXN netCostMatrix which must have positive weights/costs.
IMPORTANT: see 'TestKShortestPath.m' and 'Test graph (case 1).pdf' for netCostMatrix format.
As a follow up to Sulimon Sattari's comment. In order to accelerate the code substantially, we need to preallocate the temp array in the included dijkstra.m file. To do so replace line 31 through 38 with
temp = zeros(n,1);
for h = 1:n
if ~visited(h) % in the tree;
temp(h) = distance(h);
temp(h) = inf;
Hope this helps someone!
Thank You for the code. Is it possible to modify this so as to list the k shortest paths between all OD pairs?
Thanks for the file! I've been looking for a way to compute "all" shortest paths between two nodes and this gives me a way to do so.
I was able to speed the code up considerably (about a factor of 10) by pre-allocating "temp" in dijkstra.m, this is simply heeding MATLAB's warning to preallocate for speed. I'd recommend doing the same if anyone is bogged down with larger networks.
It cost much time in large matrix, omg
Takes a very very long time to compute if matrix is large.
Hello I'm new to Matlab, so can any one please tell me how to pass input argument for this code.?
Thank you for your sharing!
it is correct (when i run it some problem appear
Works like a charm. And as far as I could see, the only shortest path algorithm on File Exchange that returns an N-best list.