CCL

Research note

Advanced Science

Unsupervised Hierarchical Symbolic Regression for Interpretable Property Modeling in Complex Multi-Variable Systems

Siyu Lou†, Chengchun Liu†, Dongxiao Zhang, Yuntian Chen, and Fanyang Mo

A layered symbolic-learning framework that turns complex multivariable relationships into compact, chemically interpretable equations.

10.1002/advs.202521200

01 / Problem

Problem

Flexible predictive models can capture complex chemical relationships, but they often obscure which variables and combinations drive a prediction.

02 / Approach

Approach

The framework builds symbolic relationships hierarchically, combining variables into intermediate expressions and then into a compact explicit equation.

03 / Why it matters

Why it matters

Explicit equations can be inspected, challenged, and connected to chemical reasoning in ways that opaque predictors cannot, supporting both prediction and hypothesis formation.

04 / Publication details

Publication details

Authors
Siyu Lou†, Chengchun Liu†, Dongxiao Zhang, Yuntian Chen, and Fanyang Mo
Journal
Advanced Science
Year

BibTeX

@article{lou2026uhsr,
  title   = {Unsupervised Hierarchical Symbolic Regression for Interpretable Property Modeling in Complex Multi-Variable Systems},
  author  = {Siyu Lou and Chengchun Liu and Dongxiao Zhang and Yuntian Chen and Fanyang Mo},
  journal = {Advanced Science},
  year    = {2026},
  date    = {2026-01-07},
  volume  = {13},
  number  = {19},
  eid     = {e21200},
  doi     = {10.1002/advs.202521200},
  url     = {https://doi.org/10.1002/advs.202521200}
}