In this lecture, we continue to discuss multi-criteria decision making (specifically multi-objective optimization), with a shift in focus from the "Nash equilibrium" solution concept to the "Pareto equilibrium" concept. We define the "Pareto efficient set" as well as the "Pareto frontier" (and synonyms to these -- "Pareto optimal", etc.). The fundamental concept underlying Pareto efficiency is the so-called "Pareto movement", which is a topic we describe as a different way of finding decision vectors to EXCLUDE in a decision-making problem (leaving only the efficient solutions left over for the subjective consideration of a decision maker). We then pivot to describing multi-objective genetic algorithms that can be used to find the Pareto frontier/Pareto-efficient set. These algorithms adjust the notion of fitness or selection to help maintain diversity while moving a population of candidate solutions closer and closer to the Pareto frontier. Our discussion stops after introducing weighted sum methods. We will start next time with randomly weighted genetic algorithms, heading toward more modern MOGA and Pareto ranking.
Whiteboard notes for this lecture can be found at: https://www.dropbox.com/s/uxitg4gddq99dys/IEE598-Lecture3B-2022-02-08-MultiObjective_Genetic_Algorithms_MOGA_and_Friends-Part_1.pdf?dl=0
No comments:
Post a Comment