My advisor thought that my MCL algorithm behaves a bit strange, so he wanted me to investigate it. I said I'd be happy to review the code because I anyway considered refactoring, and asked if I can have another pair of eyes to help me.
A more senior PhD student was assigned to help me, and by the suggestion of my advisor we tested my code against a very well-written and well-performing implementation of MCL. This implementation was written by another professor, who is a close friend of my advisor and the actual supervisor of the student assigned to help.
But this implementation was optimized for a very specific type of maps, and on the maps I worked on it just failed consistently. The student, in a misguided attempt to protect the pride of his advisor and subsequently his, wasted days adding code and fine-tuning the implementation.
In the meantime, my MCL has a stable configuration that converges on both types of maps. It behaves differently, but the outcome is about the same as the other implementation.
I am a little sick of wasting my time (week+) on someone else's attempt to reassure their ego, so I'm doing my planned research work on the weekend...

Add Comment