Pareto Analyzer
iMOPSE includes a subproject named ParetoAnalyzer, which plays a crucial role in the analysis and comparison of multiple multi-objective optimization results across examined methods. This tool is designed to be used after conducting experiments and is thus not included in the optimizer project.
The main features of the discussed tool include:
- Calculation of True Pareto Front: In multi- and many-objective optimization, the output of each method is a set of non-dominated points. The True Pareto Front (TPF) could be defined as a set of all non-dominated solutions and can be considered the best available Pareto Front Approximation (PFA). However, in practical real-world problems, TPF is usually unknown. The Pareto Analyzer tool calculates this TPF using results generated by all runs of all compared methods.
- Nadir Point Calculation: The tool determines the Nadir Point, which is a point with the worst possible values for all objectives.
- Pareto Visualization: With the use of a Python script for the visualization part, it offers capabilities to visualize specified PFA in comparison, providing a clear and intuitive understanding of the optimization outcomes.
- Quality Measures for MO: The tool is equipped with code for the calculation of various metrics that are essential for evaluating the quality of the PFA, including Inverted Generational Distance (IGD), HyperVolume (HV), Pareto Front Size (PFS), and Purity. These measures assess convergence, diversity, volume coverage, and the proportion of non-dominated solutions in the PFA, respectively.
In summary, ParetoAnalyzer complements iMOPSE by providing tools for the analysis and evaluation of multi-objective optimization results. It facilitates visualization and quantification of the effectiveness of optimization algorithms, aiding researchers and practitioners in making informed decisions.
Python Scripts
Python scripts in iMOPSE are designed to enhance scientific research. While designed as ready-to-use tools for common tasks, they offer users the flexibility to analyze their results in a preferred way, assisting with common tasks.
- msrcpsp_solution_visualizer: Validates and visualizes MS-RCPSP solutions, aiding comprehension and refinement of optimization methods.
- multi-objective_visualizer: Elucidates trade-offs between competing objectives for multi-objective optimization researchers.
- single-objective_visualizer: Offers a graphical overview of fitness values for single-objective optimization.
- automated_experiments: Streamlines concurrent execution of iMOPSE, providing a simple way to run multiple different experiments.
These scripts enhance the efficiency and insightfulness of experiments, enriching the quality and depth of research within iMOPSE. Planned integration with the main C++ codebase will further bolster the software's robustness and scalability for broader scientific applications.