Multiobjective optimization is somewhat of a misnomer -- you actually have to have predefined weightings for each of the objectives you care about, or implement them as constraints.
0:00 - Intro
0:31 - Weighted sum method
2:37 - Pareto fronts
4:24 - Epsilon-constraint method
5:10 - Conclusion
Accompanying Python notebook:
openmdao.github.io/PracticalM...
Referenced paper for Pareto data, Brooks 2020: doi.org/10.2514/1.C035699 or www.researchgate.net/profile/...
See Chapter 9 in Engineering Design Optimization by Martins and Ning for a more in-depth view: mdobook.github.io/
Links to other relevant lectures TODO:
6 июл 2024