MToP: A MATLAB Optimization Platform for Evolutionary Multitasking

Yanchi Li, Wenyin Gong, Fei Ming, Tingyu Zhang, Shuijia Li, Qiong GuarXiv 2023 · arXiv · OpenReview

TL;DR

Researchers have developed MTOP, the first comprehensive, open-source MATLAB platform for evolutionary multitasking (EMT). MTOP provides a unified environment for evaluating, comparing, and developing new algorithms for solving multiple optimization tasks simultaneously, addressing a critical need for standardized benchmarking in this rapidly growing field.

What Problem Does This Paper Address?

Evolutionary multitasking (EMT) is a powerful paradigm that aims to solve multiple optimization problems concurrently, leveraging knowledge transfer between tasks to improve efficiency and performance. While many algorithms (MTEAs) have been proposed for EMT, the field has lacked a standardized, user-friendly platform for researchers to evaluate these algorithms, compare them against each other, and explore their application to real-world problems. This absence hinders reproducibility, makes it difficult for newcomers to enter the field, and slows down the overall progress of EMT research.

Methods

MTOP is a MATLAB-based platform designed to address this gap. Its core contributions include:

  • Comprehensive Algorithm and Problem Library: MTOP integrates over 50 MTEAs and more than 200 MTO problem instances, including both synthetic benchmarks and real-world applications. It also includes over 50 adapted single-task evolutionary algorithms to facilitate comparative studies.
  • User-Friendly Graphical Interface (GUI): The platform features an intuitive GUI with three main modules:
  • Test Module: For preliminary analysis, debugging, and qualitative visualization of algorithm performance on individual problems.
  • Experiment Module: For conducting large-scale, multi-run comparative experiments, supporting parallel execution.
  • Data Process Module: For managing, merging, and splitting experimental datasets.
  • Extensibility: MTOP is designed with an open API and modular structure, allowing users to easily add new algorithms, problems, and performance metrics.
  • Reproducibility: The platform provides extensive pre-run experimental data to enhance reproducibility and reduce computational overhead for researchers.
  • Standardized Data Structure: All experimental data is saved in a consistent .mat format, facilitating data management and post-processing.

Datasets

MTOP includes over 200 MTO problem cases, encompassing both synthetic benchmarks (e.g., CEC17-MTSO, CEC19-MTMO) and real-world applications. The paper does not introduce new benchmark datasets but rather consolidates existing ones within its platform.

Key Insights and Ideas

  • The Need for Standardization: The paper strongly emphasizes the critical need for a unified platform to advance EMT research, highlighting how the lack of standardization hinders progress.
  • MATLAB as a Strategic Choice: The authors justify their choice of MATLAB due to its prevalence in foundational EMT research, its rich ecosystem of existing code, and its strong capabilities for numerical computation and GUI development.
  • Bridging the Gap: MTOP effectively bridges the gap between algorithm development and practical evaluation by providing a comprehensive suite of tools for researchers.
  • Facilitating Comparative Analysis: The inclusion of adapted single-task EAs and a wide array of metrics allows for robust comparisons between MTEAs and traditional approaches.
  • Visualizing Complex Behaviors: The platform's visualization tools are crucial for understanding intricate phenomena like negative knowledge transfer, as demonstrated with MO-MFEA and MTDE-MKTA.

Key Results

  • Platform Functionality: MTOP successfully integrates over 50 MTEAs, 50+ single-task EAs, 200+ MTO problems, and 20+ performance metrics.
  • Reproducibility: Extensive pre-run experimental data is made available, supporting reproducibility.
  • Performance Comparison: A comparison with PlatEMO shows that MTOP has comparable computational efficiency and algorithmic implementation correctness, with no statistically significant differences in most cases for standard multi-objective algorithms.
  • User-Friendly Experience: The GUI facilitates intuitive use for testing, experimentation, and data analysis.

Limitations and Future Work

  • Potential for Bugs: The authors acknowledge that despite careful implementation, MTOP may still contain some implementation errors and bugs, and they are continuously working to identify and rectify them.
  • Future Enhancements: The authors plan to incorporate many-objective optimization codes, multitask combinatorial optimization codes, and more real-world applications in future versions. They also encourage community contributions and feedback.

Summary updated

Abstract

Evolutionary multitasking (EMT) has emerged as a popular topic of evolutionary computation over the past decade. It aims to concurrently address multiple optimization tasks within limited computing resources, leveraging inter-task knowledge transfer techniques. Despite the abundance of multitask evolutionary algorithms (MTEAs) proposed for multitask optimization (MTO), there remains a comprehensive software platform to help researchers evaluate MTEA performance on benchmark MTO problems as well as explore real-world applications. To bridge this gap, we introduce the first open-source optimization platform, named MTO-Platform (MToP), for EMT. MToP incorporates over 50 MTEAs, more than 200 MTO problem cases with real-world applications, and {over 20 performance metrics}. Moreover, to facilitate comparative analyses between MTEAs and traditional evolutionary algorithms, we adapted over 50 popular single-task evolutionary algorithms to address MTO problems. MToP boasts a user-friendly graphical interface, facilitating results analysis, data export, and schematics visualization. More importantly, MToP is designed with extensibility in mind, allowing users to develop new algorithms and tackle emerging problem domains. The source code of MToP is available at https://github.com/intLyc/MTO-Platform.
BibTeX
@misc{li2023mtop,
  author = {Yanchi Li and Wenyin Gong and Fei Ming and Tingyu Zhang and Shuijia Li and Qiong Gu},
  title = {{MToP: A MATLAB Optimization Platform for Evolutionary Multitasking}},
  year = {2023},
  eprint = {2312.08134},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2312.08134},
}
Figure