Thanks, I'll check in your improvement for k-shortest, and add epsilon
as an optional parameter for Bellman-Ford.
JVS
gu boulmi wrote:
> Hi,
>
> Concerning my contribution, meanwhile I have noticed a special case
> where the K shortest paths algorithm performs badly but fortunately I
> have found how to resolve it. That's why I send you those rectifications
> (put together with the JUnit test case).
>
> Furthermore I have some remarks about the Bellman-Ford :
>
> BellmanFordPathElement#espilon
> ---------------------------------------------------
> It would be great if it could become a user-parameter (for the moment
> the epsilon attribute is constant equal to 10-9).
> Actually, depending on the magnitude order of the cost values, the
> rouding error does not always occur at the same place beyond the comma.
> I have personally encountered cases where I had to set the epsilon to
> 10^-6 to avoid rounding errors.
> Because the epsilon value can not be changed by the user I was forced to
> create new classes (copies from BellmanFordPathElement,
> BellmanFordIterator and BellmanShortestPath) allowing this change.
>
>
> Best regards,
>
> Guillaume
|
