Interactive Evolutionary Multi-Objective Optimization via Learning-to-Rank

Interactive Evolutionary Multi-Objective Optimization via Learning-to-Rank [Paper] [Supplementary]

Overview

This repository contains implementation of the algorithm framework for Interactive Evolutionary Multi-Objective Optimization via Learning-to-Rank.

Code Structure

.
|--code --> source codes for whole project
|--log --> log files generated during execution
|--model --> parameter files for preference model
|--results --> final optimal solutions

Requirements

• C++ version: tested in C++11
• Python version: tested in Python 3.7.10
• Tensorflow version: tested in Tensorflow 2.4.0
• Operating system: tested in Ubuntu 20.04

Getting Started

Run the following script files in the folder named code:

./rebuild.sh
./run.sh

Result

The optimization results are saved in txt format. Each line in the file consists of decision variables and corresponding objective function values. They are stored under the folder:

results/out/{algorithm}/{interaction settings}/{problem}/{weight}/{seed}/

Citation

If you find our repository helpful to your research, please cite our paper:

@article{LiLY22,
    author    = {Ke Li and
                 Guiyu Lai and
                 Xin Yao},
    title     = {Interactive Evolutionary Multi-Objective Optimization via Learning-to-Rank},
    journal   = {{IEEE} Trans. Evol. Comput.},
    pages     = {1--15},
    year      = {2022},
    note      = {accepted for publication}
}