Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PymooLab: An Open-Source Visual Analytics Framework for Multi-Objective Optimization using LLM-Based Code Generation and MCDM

Thiago Santos, Sebastiao Xavier, Luiz Gustavo de Oliveira Carneiro, Gustavo de Souza

TL;DR

PymooLab is an open-source visual analytics environment built on top of \textit{pymoo} that combines visual experimentation, LLM-based modeling, and deterministic orchestration to narrow the gap between rigorous operations research and practical accessibility for domain experts.

Abstract

Multi-objective optimization is now a core paradigm in engineering design and scientific discovery. Yet mainstream evolutionary frameworks, including \textit{pymoo}, still depend on imperative coding for problem definition, algorithm configuration, and post-hoc analysis. That requirement creates a non-trivial barrier for practitioners without strong software-engineering training and often complicates reproducible experimentation. We address this gap through PymooLab, an open-source visual analytics environment built on top of \textit{pymoo}. The platform unifies configuration, execution monitoring, and formal decision support in a single reproducible workflow that automatically records hyperparameters, evaluation budgets, and random seeds. Its decoupled object-oriented architecture preserves compatibility with the base ecosystem while enabling LLM-assisted code generation for rapid model formulation. The interface also embeds interactive Multi-Criteria Decision Making (MCDM) tools, which reduces the cognitive burden of Pareto-front inspection. For computationally intensive studies, PymooLab relies on the native \textit{pymoo} acceleration pathway through JAX, improving scalability in high-dimensional evaluations. Overall, the framework combines visual experimentation, LLM-based modeling, and deterministic orchestration to narrow the gap between rigorous operations research and practical accessibility for domain experts. Source code is publicly available at https://github.com/METISBR/pymoolab.

PymooLab: An Open-Source Visual Analytics Framework for Multi-Objective Optimization using LLM-Based Code Generation and MCDM

TL;DR

PymooLab is an open-source visual analytics environment built on top of \textit{pymoo} that combines visual experimentation, LLM-based modeling, and deterministic orchestration to narrow the gap between rigorous operations research and practical accessibility for domain experts.

Abstract

Multi-objective optimization is now a core paradigm in engineering design and scientific discovery. Yet mainstream evolutionary frameworks, including \textit{pymoo}, still depend on imperative coding for problem definition, algorithm configuration, and post-hoc analysis. That requirement creates a non-trivial barrier for practitioners without strong software-engineering training and often complicates reproducible experimentation. We address this gap through PymooLab, an open-source visual analytics environment built on top of \textit{pymoo}. The platform unifies configuration, execution monitoring, and formal decision support in a single reproducible workflow that automatically records hyperparameters, evaluation budgets, and random seeds. Its decoupled object-oriented architecture preserves compatibility with the base ecosystem while enabling LLM-assisted code generation for rapid model formulation. The interface also embeds interactive Multi-Criteria Decision Making (MCDM) tools, which reduces the cognitive burden of Pareto-front inspection. For computationally intensive studies, PymooLab relies on the native \textit{pymoo} acceleration pathway through JAX, improving scalability in high-dimensional evaluations. Overall, the framework combines visual experimentation, LLM-based modeling, and deterministic orchestration to narrow the gap between rigorous operations research and practical accessibility for domain experts. Source code is publicly available at https://github.com/METISBR/pymoolab.
Paper Structure (13 sections, 4 figures, 1 table)

This paper contains 13 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Architecture and Implementation
  4. Test Module
  5. Experiment Module
  6. LLM Prompt Agent
  7. Extensibility Module
  8. MCDM Module
  9. Operational Usability and Workflow Demonstration
  10. Empirical Benchmarking and Statistical Synthesis
  11. Low-Overhead Formulation via the LLM Agent
  12. A Posteriori Selection (MCDM)
  13. Conclusion

Figures (4)

  • Figure 1: System topography of the framework. The architecture maps the epistemological lifecycle of an optimization study: translating domain knowledge via LLMs, configuring empirical parameters, executing deterministic trials, and extracting compromise solutions via MCDM.
  • Figure 2: Test Module interface of PymooLab. The module organizes single-run validation into six coordinated regions: algorithm selection (A), problem selection (B), parameter and backend configuration (C), metric selection (D), result display and execution controls (E), and execution log for traceability (F). Finalized runs are also persisted as structured payloads for subsequent reload and analysis.
  • Figure 3: Automated analytical reporting within the Experiment Module. The interface natively compiles multi-run indicator statistics, substantially accelerating the benchmarking publication pipeline.
  • Figure 4: A posteriori decision analysis in the integrated Analysis & MCDM workspace. The framework identifies compromise solutions directly over persisted Pareto manifolds while mapping decision traceability through auditable sidecar records.