Open mitchwagner opened 6 years ago
Try setting --edge-penalty to a large number, e.g., the number of edges in the graph.
The resulting paths seem to be in the same order when I use 21 as the --edge-penalty parameter, though the weights still do not match. The weights I obtain are much smaller, whereas the example weights are all greater than 1.
Did we change the code at some point to invert the log transform of the edge weights before printing path costs?
That might be it. The change you mention appears to have been added in c6ab78eea8c5ee44a13cc5396d2b203e801dc5cc, which is subsequent to the sample files being added.
Running the example output using the current example input and the current code does not seem to produce the same output (specifically, the weights for the paths seem to be different). This issue extends to before commit 0b4f105d162d79409037cb40641a9b5c1c63cd5d.
I attempted running it with both --PageRank and with the pre-weighted input.