Software

qPOTS is an open-source Python package from Penn State’s Computational Complex Engineered Systems Design Laboratory (CSDL) for efficient batch (parallel) multiobjective Bayesian optimization of expensive black-box problems. Its core method—batch Pareto Optimal Thompson Sampling (qPOTS)—fits Gaussian-process surrogate models for each objective (and optional constraints), draws posterior sample paths, and selects a diverse batch of candidate designs from the resulting predicted Pareto front, sidestepping the brittle “maximize a complex acquisition function” inner-loop that often dominates MOBO runtimes. Built on the PyTorch/BoTorch ecosystem, qPOTS includes utilities for model fitting and hypervolume tracking, plus reference implementations of common MOBO baselines (e.g., ParEGO and entropy-based methods such as PESMO/MESMO/JESMO) and ready-to-run constrained and unconstrained examples. The package is available on PyPI (pip install qpots) with full online documentation and is released under a GPL-3.0 license. 

The theory behind qPOTS is published in the proceedings of AISTATS 2025: https://proceedings.mlr.press/v258/renganathan25a.html