A graphical front end and analysis tool for VAMOS, a Python framework for multi-objective optimization. It is inspired by Evolver-Studio, with VAMOS as its backend.
Status: first prototype (0.1.0.dev0): NSGA-II on the real-coded ZDT and DTLZ problems (DTLZ with two or three objectives, drawn in 3D). The other algorithms, encodings and problems are listed in the app as coming soon.
| Section | Page | |
|---|---|---|
| Explore | Algorithms | The parameter space of each algorithm as a grammar: tree, table, default configuration and YAML |
| Problems | The problem families, with each problem's arguments, bounds, reference front (in 3D for three objectives) and the approximate fronts built from runs | |
| Quality indicators | What each indicator measures, with a sandbox to see how they respond | |
| Optimize | New run | Choose a problem and set its arguments, choose an algorithm and adjust its configuration, set the budget and launch one run or a series of independent runs |
| Monitor | Follow a run or a series live, with the fronts as they converge; stop it keeping its results, or cancel it | |
| History | Every run with its state; open one in the monitor or the results | |
| Analyze | Results | The front with its knee and extremes, the quality indicators, the solutions and the details of the run; for a series of independent runs, their fronts and union and their quality indicators with summary statistics, or any of its runs |
| Learn | Tutorials | Interactive tutorials by level (S1–S4 basic so far): runs launched from the tutorial with the values to expect, and Jupyter notebooks that do the same with VAMOS's Python API |
Runs go on in the background and are kept in solve-runs/ (set VAMOS_STUDIO_RUNS to use another
directory), each with VAMOS's run artifact. For problems without a known reference front, runs can
build an approximate one, kept in solve-runs/reference-fronts/. Fronts are measured against jMetal's reference fronts,
the same files Evolver-Studio uses, so the values of both tools match.
VAMOS-Studio needs Python 3.11 or newer. With Conda:
make env # creates (or updates) the vamos-studio Conda environment
make run # opens the app in the browserOr with any virtual environment:
python -m pip install -e ".[dev]"
streamlit run app.pyInstall the fast extra (pip install -e ".[fast]") for VAMOS's Numba kernels.
make test # pytest
make lint # ruff check and format check
make format # ruff formatSee CLAUDE.md for the architecture, CODING_GUIDELINES.md and GIT_GUIDELINES.md.
VAMOS-Studio is distributed under the MIT license.