Search papers, labs, and topics across Lattice.
This paper introduces Element-Weighted Kolmogorov-Arnold Networks (KANs) for predicting crystal properties, achieving state-of-the-art accuracy in predicting formation energy, band gap, and work function. KANs, unlike traditional neural networks, use learnable activation functions, enabling the extraction of interpretable chemical trends. Through embedding analysis, correlation studies, and PCA, the authors demonstrate that KANs can uncover relationships aligned with the periodic table and quantum mechanical principles without explicit physical constraints.
KANs don't just predict crystal properties with SOTA accuracy; they also reveal the underlying chemistry, offering a glimpse into the model's reasoning.
Characterizing crystalline energy landscapes is essential to predicting thermodynamic stability, electronic structure, and functional behavior. While machine learning (ML) enables rapid property predictions, the "black-box" nature of most models limits their utility for generating new scientific insights. Here, we introduce Kolmogorov-Arnold Networks (KANs) as an interpretable framework to bridge this gap. Unlike conventional neural networks with fixed activation functions, KANs employ learnable functions that reveal underlying physical relationships. We developed the Element-Weighted KAN, a composition-only model that achieves state-of-the-art accuracy in predicting formation energy, band gap, and work function across large-scale datasets. Crucially, without any explicit physical constraints, KANs uncover interpretable chemical trends aligned with the periodic table and quantum mechanical principles through embedding analysis, correlation studies, and principal component analysis. These results demonstrate that KANs provide a powerful framework with high predictive performance and scientific interpretability, establishing a new paradigm for transparent, chemistry-based materials informatics.
