Progressively Interactive Evolutionary Multi-Objective Optimization ...
Progressively Interactive Evolutionary Multi-Objective Optimization ...
Progressively Interactive Evolutionary Multi-Objective Optimization ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
An <strong>Interactive</strong> <strong>Evolutionary</strong> <strong>Multi</strong>-<strong>Objective</strong><br />
<strong>Optimization</strong> Method Based on Polyhedral Cones<br />
Ankur Sinha, Pekka Korhonen, Jyrki Wallenius and Kalyanmoy Deb ⋆⋆<br />
Department of Business Technology<br />
Aalto University School of Economics<br />
PO Box 21220, 00076 Aalto, Helsinki, Finland<br />
{Firstname.Lastname}@hse.fi<br />
Abstract. This paper suggests a preference based methodology, where the information<br />
provided by the decision maker in the intermediate runs of an evolutionary<br />
multi-objective optimization algorithm is used to construct a polyhedral cone.<br />
This polyhedral cone is used to eliminate a part of the search space and conduct<br />
a more focussed search. The domination principle is modified, to look for better<br />
solutions lying in the region of interest. The search is terminated by using a local<br />
search based termination criterion. Results have been presented on two to five<br />
objective problems and the efficacy of the procedure has been tested.<br />
Keywords: <strong>Evolutionary</strong> multi-objective optimization, multiple criteria decisionmaking,<br />
interactive multi-objective optimization, sequential quadratic programming,<br />
preference based multi-objective optimization.<br />
1 Introduction<br />
Most of the existing evolutionary multi-objective optimization (EMO) algorithms aim<br />
to find a set of well-converged and well-diversified Pareto-optimal solutions [1, 2]. As<br />
discussed elsewhere [3, 5], finding the entire set of Pareto-optimal solutions has its own<br />
intricacies. Firstly, the usual domination principle allows a majority of the population<br />
members to become non-dominated, thereby not allowing much room for introducing<br />
new solutions in a finite population. This slows down the progress of an EMO algorithm.<br />
Secondly, the representation of a high-dimensional Pareto-optimal front requires<br />
an exponentially large number of points, thereby requiring a large population size in<br />
running an EMO procedure. Thirdly, the visualization of a high-dimensional front becomes<br />
a non-trivial task for decision-making purposes.<br />
In most of the existing EMO algorithms the decision maker is usually not involved<br />
during the optimization process. The decision maker is called only at the end of the<br />
optimization run after a set of approximate Pareto-optimal solutions has been found.<br />
The decision making process is then executed by choosing the most preferred solution<br />
from the set of approximate Pareto-optimal solutions obtained. This approach is called<br />
⋆⋆ Also Department of Mechanical Engineering, Indian Institute of Technology Kanpur, PIN<br />
208016, India (deb@iitk.ac.in).<br />
58