My optimization suite of choice these days is mystic, an open-source framework written in Python. As a whole, mystic is a very general and flexible optimization framework, with swappable modular optimization strategies, termination criteria, constraints tools, and so so, not tied to any particular problem or application.
For me, specifically, mystic is the workhorse behind the current implementation of the Optimal Uncertainty Quantification (OUQ) project. In OUQ, we’re interested in calculating optimal bounds in uncertain outputs given some information about uncertain inputs. Those pieces of information are essentially constraints: the moral is that any admissible candidate for some uncertain reality is constrained to be consistent with your information about that reality. So, there’s a natural need for constrained global optimization tools — mystic fits the bill nicely.
Along with mystic’s primary maintainer, Mike McKerns, and a summer student, Lan Huong Nguyen, I’ll be making some contributions to mystic’s OUQ toolset over the next few months — news to follow as it comes.