In this lecture, we finish our introduction to multi-objective genetic algorithms (MOGA) by reviewing weighted sum approaches (like RWGA), alternating objective approaches (like VEGA), and then introducing Pareto-ranking approaches (like MOGA and NSGA-2). We discuss how Pareto-ranking approaches use a ranking selection approach where fitness is tied to how many other solutions either dominate the focal solution or are dominated by it. That also allows us to introduce basic fitness sharing (which will be revisited as we move forward to discuss multi-modal optimization in the next section). Whereas Pareto ranking ensures population movement toward the Pareto frontier, fitness sharing (and related measures) help ensure movement along the frontier to sample a larger fraction of it in the terminal population.
Whiteboard notes for this lecture can be found at: https://www.dropbox.com/s/2tvix28ylq9gmiw/IEE598-Lecture3C-2022-02-10-MultiObjective_Genetic_Algorithms_MOGA_and_Friends-Part_2.pdf?dl=0
No comments:
Post a Comment