SciPost Submission Page
Rotation-equivariant Graph Neural Networks for Learning Glassy Liquids Representations
by Francesco Saverio Pezzicoli, Guillaume Charpiat, François P. Landes
This is not the latest submitted version.
This Submission thread is now published as
Submission summary
| Authors (as registered SciPost users): | François P. Landes |
| Submission information | |
|---|---|
| Preprint Link: | scipost_202310_00011v2 (pdf) |
| Code repository: | https://doi.org/10.5281/zenodo.10805522 |
| Data repository: | https://doi.org/10.5281/zenodo.10805522 |
| Date submitted: | 2024-03-13 11:01 |
| Submitted by: | Landes, François P. |
| Submitted to: | SciPost Physics |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approaches: | Theoretical, Computational |
Abstract
The difficult problem of relating the static structure of glassy liquids and their dynamics is a good target for Machine Learning, an approach which excels at finding complex patterns hidden in data. Indeed, this approach is currently a hot topic in the glassy liquids community, where the state of the art consists in Graph Neural Networks (GNNs), which have great expressive power but are heavy models and lack interpretability. Inspired by recent advances in the field of Machine Learning group-equivariant representations, we build a GNN that learns a robust representation of the glass' static structure by constraining it to preserve the roto-translation (SE(3)) equivariance. We show that this constraint significantly improves the predictive power at comparable or reduced number of parameters but most importantly, improves the ability to generalize to unseen temperatures. While remaining a Deep network, our model has improved interpretability compared to other GNNs, as the action of our basic convolution layer relates directly to well-known rotation-invariant expert features. Through transfer-learning experiments displaying unprecedented performance, we demonstrate that our network learns a robust representation, which allows us to push forward the idea of a learned structural order parameter for glasses.
Author comments upon resubmission
We have replied to the referees.
List of changes
We upload a pdf with changes highlighted in blue for convenience.
Current status:
Reports on this Submission
Report #2 by Joerg Rottler (Referee 1) on 2024-3-28 (Invited Report)
Report
I appreciate the authors' thoughtful responses to my first report. The manuscript has also been improved in multiple places and can now be accepted. It would have been nice to see some of the authors' responses added to the main text so that other readers also benefit. Right now, only absolutely required modifications were made. They could still take up that opportunity when they submit their final version should they so decide.
Report
The authors have answered in detail to the critique and have revised their manuscript accordingly.
Since the method proposed in the manuscript opens a new pathway in learning structure-dynamics correlations in supercooled liquids with interesting applications for better understanding dynamic heterogeneity, I recommend publication of "Rotation-equivariant Graph Neural Networks for Learning Glassy Liquids Representations" in SciPost Physics.