A. Romero

Chowdhury Mohammad Abid Rahman, Ghadendra Bhandari, Nasser M Nasrabadi et al.

Machine learning (ML) models have emerged as powerful tools for accelerating materials discovery and design by enabling accurate predictions of properties from compositional and structural data. These capabilities are vital for developing advanced technologies across fields such as energy, electronics, and biomedicine, potentially reducing the time and resources needed for new material exploration and promoting rapid innovation cycles. Recent efforts have focused on employing advanced ML algorithms, including deep learning - based graph neural network, for property prediction. Additionally, ensemble models have proven to enhance the generalizability and robustness of ML and DL. However, the use of such ensemble strategies in deep graph networks for material property prediction remains underexplored. Our research provides an in-depth evaluation of ensemble strategies in deep learning - based graph neural network, specifically targeting material property prediction tasks. By testing the Crystal Graph Convolutional Neural Network (CGCNN) and its multitask version, MT-CGCNN, we demonstrated that ensemble techniques, especially prediction averaging, substantially improve precision beyond traditional metrics for key properties like formation energy per atom ($ΔE^{f}$), band gap ($E_{g}$) and density ($ρ$) in 33,990 stable inorganic materials. These findings support the broader application of ensemble methods to enhance predictive accuracy in the field.

Charles Rhys Campbell, Aldo H. Romero, Kamal Choudhary

Generative models have become significant assets in the exploration and identification of new materials, enabling the rapid proposal of candidate crystal structures that satisfy target properties. Despite the increasing adoption of diverse architectures, a rigorous comparative evaluation of their performance on materials datasets is lacking. In this work, we present a systematic benchmark of three representative generative models- AtomGPT (a transformer-based model), Crystal Diffusion Variational Autoencoder (CDVAE), and FlowMM (a Riemannian flow matching model). These models were trained to reconstruct crystal structures from subsets of two publicly available superconductivity datasets- JARVIS Supercon 3D and DS A/B from the Alexandria database. Performance was assessed using the Kullback-Leibler (KL) divergence between predicted and reference distributions of lattice parameters, as well as the mean absolute error (MAE) of individual lattice constants. For the computed KLD and MAE scores, CDVAE performs most favorably, followed by AtomGPT, and then FlowMM. All benchmarking code and model configurations will be made publicly available at https://github.com/atomgptlab/atombench_inverse.

Chowdhury Mohammad Abid Rahman, Aldo H. Romero, Prashnna K. Gyawali

Accurate prediction of material properties facilitates the discovery of novel materials with tailored functionalities. Deep learning models have recently shown superior accuracy and flexibility in capturing structure-property relationships. However, these models often rely on supervised learning, which requires large, well-annotated datasets an expensive and time-consuming process. Self-supervised learning (SSL) offers a promising alternative by pretraining on large, unlabeled datasets to develop foundation models that can be fine-tuned for material property prediction. In this work, we propose supervised pretraining, where available class information serves as surrogate labels to guide learning, even when downstream tasks involve unrelated material properties. We evaluate this strategy on two state-of-the-art SSL models and introduce a novel framework for supervised pretraining. To further enhance representation learning, we propose a graph-based augmentation technique that injects noise to improve robustness without structurally deforming material graphs. The resulting foundation models are fine-tuned for six challenging material property predictions, achieving significant performance gains over baselines, ranging from 2% to 6.67% improvement in mean absolute error (MAE) and establishing a new benchmark in material property prediction. This study represents the first exploration of supervised pertaining with surrogate labels in material property prediction, advancing methodology and application in the field.