We introduce the multitask optimization platform, named MToP, for evolutionary multitasking:
- 50+ multitask evolutionary algorithms for multitask optimization
- 50+ single-task evolutionary algorithms that can handle multitask optimization problems
- 200+ multitask optimization problem cases with real-world applications
- 150+ classical single-task optimization benchmark problems
- 20+ performance metrics covering single- and multi-objective optimization
MToP is a user-friendly tool with a graphical user interface that makes it easy to analyze results, export data, and plot schematics. More importantly, MToP is extensible, allowing users to develop new algorithms and define new problems.
📄 Documents: [Paper] / [User Guide]
💿 Data: In order to enhance reproducibility and avoid unnecessary repeated experiments, we provide the experimental data of algorithms on benchmark suites over 30 independent runs under fixed random seeds: [Google Drive]
Copyright (c) Yanchi Li. You are free to use the MToP for research purposes. All publications which use this platform should acknowledge the use of "MToP" or "MTO-Platform" and cite as "Y. Li, W. Gong, F. Ming, T. Zhang, S. Li, and Q. Gu, MToP: A MATLAB Optimization Platform for Evolutionary Multitasking, 2023, arXiv:2312.08134"
@Article{Li2023MToP,
title = {{MToP}: A {MATLAB} Optimization Platform for Evolutionary Multitasking},
author = {Yanchi Li and Wenyin Gong and Fei Ming and Tingyu Zhang and Shuijia Li and Qiong Gu},
journal = {arXiv preprint arXiv:2312.08134},
year = {2023},
eprint = {2312.08134},
}
Email: int_lyc@cug.edu.cn
QQ Group: 862974231
New Algorithms: 🆕
- DTSKT (Single-objective Many-task TSMC 25)
- MTEA-PAE (Single-objective Multi-task ASOC 25)
- MO-MTEA-PAE (Multi-objective Multi-task ASOC 25)
- EMTO-AI (Single-objective Multi-task TETCI 24)
New Problems: 📊
- STOP-G (Many-task optimization problems TCYB 2025)
New Features: ✨
- Global random seed control using rng(seed + rep -1) for better reproducibility
- Add IOHanalyzer data type export to support more analysis tools
- Add option to save best decision variables during optimization
- Add competitive and constrained buttons to Experiment module for enhanced problem type support
- Add context menus component for batch editing of problem public parameters with dedicated editing dialog (thanks @Zhizhou-Geng)
- Integrate refresh functionality into right-click context menu with keyboard shortcuts
Optimizations and Improvements: ⚡
- Refactor constraint handling logic with full constraint support
- Add Evaluation method to Problem class for flexible evaluation
- Improve GUI display and user interface responsiveness
- Change relative data load path for improved file management
- Improve Experiment module buttons display and ordering
- Update CMA-ES related algorithms to follow official implementation
- Algorithm reloading now triggers automatically when clicking Add button
Older Release Highlights (Click to Expand)
- New Algorithm: SSLT (Single-objective Multi-task KBS-25)
- New Problems: MGA_DSM_GTOP and MGA-GTOP (Multi-task Global Trajectory Optimization Problem)
- Add the Friedman test for statistical analysis
- Add Convergence range drawing with 95% confidence interval
- Improve the GUI tooltip and description
- Fix some bugs
- New Algorithms:
- RVC-MTEA (Competitive Multi-objective Multi-task TEVC-24)
- MTEA-DCK (Multi-objective Multi-task TSMC-S-25)
- MFEA-VC (Single-objective Multi-task ASOC-24)
- CMO-LKT (Constrained Single-objective Multi-task TSMC-S-25)
- New Problems:
- CMOMT Benchmark (Competitive Multi-objective Multi-task TEVC-24)
- MOSCP2 (Competitive Multi-objective Sensor Coverage Problem)
- OPF-CMOMT (Competitive Multi-objective Optimal Power Flow)
- Optimize 2D Pareto Front drawing
- Add competitive multi-objective multi-task metric IGD-CMT and HV-CMT
- Update MTS metric with convergence for HV, IGD, and IGD+
- Fix MFEA-GHS domain adaptation bug
- Fix LDA-MFEA data size reduce method
- Fix NaN bug in IGD and IGD+ calculation
- Fix the bug of multifactorial algorithms run in many-task problems
- New Algorithm:
- TNG-NES (Single-objective Many-task TEVC24)
- MTDE-ADKT (Single-objective Multi-task ASOC24)
- AR-MOEA, MSEA (Multi-objective Single-task)
- New Problem: LSMaTSO (Large-scale many-task single-objective)
- Fix the bug when GUI parallel runs experiments with save Dec.
- New Algorithm: MTEA-HKTS (Single-objective Multi/Many-task INS24)
- New Problem: Multi-objective sensor coverage problem
- New features:
- Draw dynamic Dec and Obj of populations during optimization in the Test Module
- Pause and Stop buttons can now respond in time by clicking on both the Test and Experiment Module
- Figures sample numbers in the Test Module can be modified, and figures can be exported
- Algorithm and Problem objects can be input in the command line running e.g. "mto(MFEA(), CMT1());"
- New Algorithms:
- CEDA (Constrained Single-objective Multitask SWEC24)
- MTEA-D-TSD (Multi-objective Multitask GECCO24)
- Global-GA (Single-objective Single-task TEVC24)
- KLDE and KLPSO (Single-objective Single-task TEVC23)
- Other classical algorithms: RVEA (MO-ST), SMS-EMOA (MO-ST), IPOP-CMA-ES (SO-ST)
- New Problems:
- Classical Single-Objective Functions with any dimension setting
- Fix some bugs.
- Newly added algorithms:
- MTDE-MKTA (multi-objective multitask TEVC 2024) with application problems
- KR-MTEA (multi/single-objective multitask INS 2023)
- Fix some bugs.
- Newly added algorithms:
- TRADE (single-objective many-task TCYB 2023)
- ASCMFDE (single-objective multitask TEVC 2021)
- Add error value type of WCCI20-MTSO
- Update Operator GA (SBX and polynomial mutation) with more advanced calculation methods. GA-based algorithms now have improved performance.
- The speed of experimental execution is significantly increased, brought by the simultaneous evaluation of whole population decision variables
- 3D task figures of 2-dimensional variables for un-/constrained single-objective multi-/many-/single-task optimization can be plotted in the test module
- Performance metrics can be displayed automatically based on the data type in the experiment module
- Newly added algorithms:
- MKTDE (single-objective multi-task TEVC 2022)
- CCEF-ECHT (constrained single-objective TSMC 2023)