MToP: A MATLAB Benchmarking Platform for Evolutionary Multitasking

TL;DR

This work introduces MToP, an open-source MATLAB benchmarking platform tailored for evolutionary multitasking (EMT) and multitask optimization (MTO). It provides over 50 MTEAs, 200+ benchmark MTO problems, 20+ metrics, and a GUI-driven workflow plus a command-line interface, enabling reproducible large-scale experiments and comprehensive analyses. By offering standardized data structures (MTOData.mat), modular code patterns, and extensive pre-run data, MToP facilitates fair comparisons between MTEAs and traditional EAs, supports exploratory problem analysis, and promotes extensibility for future EMT developments. Empirical validation demonstrates reproducibility and competitive performance relative to PlatEMO, highlighting MToP’s potential to accelerate EMT research and real-world applications. The platform emphasizes reproducibility, data management, and visualization to aid researchers in diagnosing knowledge transfer dynamics and in selecting appropriate benchmarks and algorithms for specific MTO scenarios.

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 need for 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 benchmarking platform, named MToP, for EMT. MToP incorporates over 50 MTEAs, more than 200 MTO problem cases with real-world applications, and over 20 performance metrics. Based on these, we provide benchmarking recommendations tailored for different MTO scenarios. Moreover, to facilitate comparative analyses between MTEAs and traditional evolutionary algorithms, we adapted over 50 popular single-task evolutionary algorithms to address MTO problems. Notably, we release extensive pre-run experimental data on benchmark suites to enhance reproducibility and reduce computational overhead for researchers. MToP features a user-friendly graphical interface, facilitating results analysis, data export, and schematic 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

Figures (12)

• Figure 1: Examples of graphical display in the Test Module of MToP. (a) and (b) illustrate the landscapes of single-objective problems with different tasks in the one- and two-dimensional unified search space, respectively. (c) shows the feasible regions of a single-objective problem with different tasks in the two-dimensional unified search space. (d) depicts the Pareto front of a multi-objective problem with multiple tasks. (e) displays the convergence behavior of metrics after executing algorithms on problems.
• Figure 2: schematic plotting by the Experiment Module of MToP. (a) Convergence plot of algorithms on problems. (b) Pareto front plot of multi-objective problems.
• Figure 3: Class diagram of MToP. The main classes include Algorithm, Problem, Individual.
• Figure 4: Sequence diagram of MToP. The main workflow involves MTO_GUI / MTO_CMD, Algorithm, Problem, and Metric.
• Figure 5: File structure of MToP. The root directory contains the main script file mto.m and four subfolders: Algorithms/, Problems/, Metrics/, and GUI/.