Bgolearn

Bgolearn

A unified Bayesian optimization framework for accelerating materials discovery.

Language: English | 简体中文 | 日本語 | 한국어 | Deutsch

Playground: Interactive Bayesian optimization game

---

📺 Media Coverage

Featured on Dragon TV · Shanghai Media Group (SMG)

Our work was featured on Dragon TV (东方卫视), highlighting its research and applications.

_{Featured on Dragon TV (东方卫视), Shanghai Media Group (SMG).}

---

Featured Introduction

Bayesian Global Optimization is Chapter 1 of the Springer book An Introduction to Materials Informatics by Prof. Tong-Yi Zhang, Academician of the Chinese Academy of Sciences. The active-learning examples and results in this chapter are implemented with and depend on Bgolearn.

Overview

Bgolearn is a research-oriented Python framework for Bayesian Global Optimization (BGO). It is designed for data-driven materials discovery, experimental design, and virtual screening, where each new measurement can be costly and every recommendation should be traceable.

The framework brings surrogate modeling, uncertainty-aware acquisition, active learning, and candidate ranking into a single workflow. It supports both regression and classification tasks, so researchers can move from small experimental datasets to the next most informative material candidates with less custom glue code.

Highlights

• Unified workflows for regression, classification, active learning, and virtual screening.
• Multiple surrogate models, including Gaussian process, SVM, random forest, AdaBoost, and MLP-based options.
• Acquisition functions for noisy and noise-free optimization, including EI, AEI, EQI, REI, UCB, PoI, PES, and KG.
• Candidate recommendation for minimization, maximization, and multi-objective research workflows.
• A lightweight local web interface for users who prefer an interactive workflow.

Resources

Resource	Link
Paper	npj Computational Materials
Book chapter	Bayesian Global Optimization
Manual	bgolearn.netlify.app
Video tutorial	Bilibili
Conference report	CMC 2025
Multi-object module	MultiBgolearn
Official GUI	BgoFace
Example code and data	CodeDemo

Representative Applications

Bgolearn has supported a growing set of materials science and engineering studies. The following papers are a representative subset:

2026

Venue	Application
Surfaces and Interfaces	Photochemical synthesis of WS2 thin films
JPhys Materials	Design of TaNbMoVW refractory high-entropy alloys
Chemical Science	Optimizing on-surface reactions
Springer Nature	Book chapter: Bayesian Global Optimization
Journal of Materials Informatics	Optimizing SciBERT hyperparameters
Science Bulletin	Discovering ultra-durable and highly active catalysts
Aggregate	Discovering G-quadruplex deep-eutectic circularly polarized luminescence materials
Materials Science in Semiconductor Processing	Substitutional Hf-doped p-type MoS2 via pulsed-laser synthesis

2025

Venue	Application
Nano Letters	Self-driving laboratory under UHV
Small	ML-engineered nanozyme system for anti-tumor therapy
Computational Materials Science	Mg-Ca-Zn alloy optimization
Measurement	Foaming agent optimization in EPB shield construction
Intelligent Computing	Metasurface design via Bayesian learning

2024

Venue	Application
Materials & Design	Lead-free solder alloys via active learning
npj Computational Materials	MLMD platform with Bgolearn backend

Installation

Install from PyPI:

pip install Bgolearn

Upgrade to the latest release:

pip install --upgrade Bgolearn

Check the installed version:

pip show Bgolearn

Run the Interface

Clone the repository and start the local UI:

git clone https://github.com/Bin-Cao/Bgolearn.git
cd Bgolearn
python bgolearn_ui.py

Then open:

http://127.0.0.1:8787

Citation

If Bgolearn supports your research, please cite:

@article{Cao2026Bgolearn,
  author    = {Bin Cao and Jie Xiong and Jiaxuan Ma and Yuan Tian and Yirui Hu and Mengwei He and Longhan Zhang and Jiayu Wang and Jian Hui and Li Liu and Dezhen Xue and Turab Lookman and Jun Wang and Tong-Yi Zhang},
  title     = {Bgolearn: a unified Bayesian optimization framework for accelerating materials discovery},
  journal   = {npj Computational Materials},
  year      = {2026},
  doi       = {10.1038/s41524-026-02226-3},
  issn      = {2057-3960},
  publisher = {Springer Nature},
  url       = {https://doi.org/10.1038/s41524-026-02226-3}
}

Funding

Bgolearn was selected for the Open-Source Artificial Intelligence Support Program (2025) supported by the Shanghai Municipal Commission of Economy and Informatization.

Project material: figures/funding.png

License

Bgolearn is released under the MIT License.