Closed nightdr closed 3 years ago
I will also email you and Jake some profiling info on the algorithm after I document it. From testing on my setup, this version of the algorithm is 20% - 30% quicker than the old one when benchmarked using ParallelArchipelago, ExplicitRegression, etc. (you can see the deleted fork_mutation_benchmarks.py for the file I used to benchmark the algorithms). Furthermore, I will test the algorithm on Jake's problem to see how much of the GPSR process is spent in the fork mutation.
Brought fork mutation algorithm from the fork mutation branch into develop and refactored the algorithm to be easier to maintain and more optimized. To stick to my promise that I would PR today (Friday), I am making the PR right now, but I plan to comment the algorithm in the near future (next day or two) to make it easier to understand.