There are 3 algorithm that should do nearly the same but they are different in order to improve performance.
Forst of all NW: [url removed, login to view]
Some A* variation of it is used in YA : [url removed, login to view] // need analyse python source code
And also there is GPU version of it GP: [url removed, login to view]
In YA they did not use normal NW in order improve speed of the program. But what I discovered is that if we run YA that uses normal NW insted of their optimized A* method I get better quality results. I need to know why is it possible. What is more what I use GP algorithm instead of YA I get faster results and those results are even better thatn those from NW.
So what I need is a comparison (description + diagrams with description and selection in dofferences) of those 3 algorithms. I also need to know why (basing on the algorithm) it is possible for them to differ in results and why one can possibly retursn better results than the other.