publications (ML/GNN)

2025

1. ML/GNN

   The Open Molecules 2025 (OMol25) Dataset, Evaluations, and Models

   Daniel S. Levine, Muhammed Shuaibi, Evan Walter Clark Spotte-Smith, and 20 more authors

   arXiv preprint arxiv:2505.08762, 2025
2. ML/GNN

   All-atom Diffusion Transformers: Unified generative modelling of molecules and materials

   Chaitanya K. Joshi, Xiang Fu, Yi-Lun Liao, and 4 more authors

   arXiv preprint arxiv:2503.03965, 2025

2024

1. ML/GNN

   From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction

   Nima Shoghi, Adeesh Kolluru, John R Kitchin, and 3 more authors

   arXiv preprint arXiv:2310.16802, 2024
2. ML/GNN

   The Open DAC 2023 Dataset and Challenges for Sorbent Discovery in Direct Air Capture

   Anuroop Sriram, Sihoon Choi, Xiaohan Yu, and 6 more authors

   ACS Central Science, 2024

   DOI
3. ML/GNN

   Fine-Tuned Language Models Generate Stable Inorganic Materials as Text

   Nate Gruver, Anuroop Sriram, Andrea Madotto, and 3 more authors

   ICLR, 2024
4. ML/GNN

   Generalization of graph-based active learning relaxation strategies across materials

   Xiaoxiao Wang, Joseph Musielewicz, Richard Tran, and 6 more authors

   Machine Learning: Science and Technology, 2024

2023

1. ML/GNN

   Beyond independent error assumptions in large GNN atomistic models

   Janghoon Ock, Tian Tian, John Kitchin, and 1 more author

   The Journal of Chemical Physics, 2023
2. ML/GNN

   AmpTorch: A Python package for scalable fingerprint-based neural network training on multi-element systems with integrated uncertainty quantification

   Muhammed Shuaibi, Yuge Hu, Xiangyun Lei, and 8 more authors

   Journal of Open Source Software, 2023
3. ML/GNN

   Chemical Properties from Graph Neural Network-Predicted Electron Densities

   Ethan M Sunshine, Muhammed Shuaibi, Zachary W Ulissi, and 1 more author

   The Journal of Physical Chemistry C, 2023

2022

1. ML/GNN

   How Do Graph Networks Generalize to Large and Diverse Molecular Systems?

   Johannes Gasteiger, Muhammed Shuaibi, Anuroop Sriram, and 4 more authors

   arXiv preprint arXiv:2204.02782, 2022

   DOI
2. ML/GNN

   Transfer learning using attentions across atomic systems with graph neural networks (TAAG)

   Adeesh Kolluru, Nima Shoghi, Muhammed Shuaibi, and 4 more authors

   The Journal of Chemical Physics, 2022

   DOI
3. ML/GNN

   Spherical Channels for Modeling Atomic Interactions

   C. Lawrence Zitnick, Abhishek Das, Adeesh Kolluru, and 5 more authors

   NeurIPS, Dec 2022

   DOI
4. ML/GNN

   Robust and scalable uncertainty estimation with conformal prediction for machine-learned interatomic potentials

   Yuge Hu, Joseph Musielewicz, Zachary W Ulissi, and 1 more author

   Machine Learning: Science and Technology, Dec 2022

   Abs DOI

   Uncertainty quantification (UQ) is important to machine learning (ML) force fields to assess the level of confidence during prediction, as ML models are not inherently physical and can therefore yield catastrophically incorrect predictions. Established a-posteriori UQ methods, including ensemble methods, the dropout method, the delta method, and various heuristic distance metrics, have limitations such as being computationally challenging for large models due to model re-training. In addition, the uncertainty estimates are often not rigorously calibrated. In this work, we propose combining the distribution-free UQ method, known as conformal prediction (CP), with the distances in the neural network’s latent space to estimate the uncertainty of energies predicted by neural network force fields. We evaluate this method (CP+latent) along with other UQ methods on two essential aspects, calibration, and sharpness, and find this method to be both calibrated and sharp under the assumption of independent and identically-distributed (i.i.d.) data. We show that the method is relatively insensitive to hyperparameters selected, and test the limitations of the method when the i.i.d. assumption is violated. Finally, we demonstrate that this method can be readily applied to trained neural network force fields with traditional and graph neural network architectures to obtain estimates of uncertainty with low computational costs on a training dataset of 1 million images to showcase its scalability and portability. Incorporating the CP method with latent distances offers a calibrated, sharp and efficient strategy to estimate the uncertainty of neural network force fields. In addition, the CP approach can also function as a promising strategy for calibrating uncertainty estimated by other approaches.

2021

1. ML/GNN

   Rotation Invariant Graph Neural Networks using Spin Convolutions

   Muhammed Shuaibi, Adeesh Kolluru, Abhishek Das, and 4 more authors

   arXiv preprint arXiv:2106.09575, Dec 2021

   DOI
2. ML/GNN

   The Open Catalyst Challenge 2021: Competition Report.

   Abhishek Das, Muhammed Shuaibi, Aini Palizhati, and 8 more authors

   In NeurIPS (Competition and Demos), Dec 2021

2020

1. ML/GNN

   Methods for comparing uncertainty quantifications for material property predictions

   Kevin Tran, Willie Neiswanger, Junwoong Yoon, and 3 more authors

   Machine Learning: Science and Technology, May 2020

   DOI

2019

1. ML/GNN

   Convolutional Neural Network of Atomic Surface Structures To Predict Binding Energies for High-Throughput Screening of Catalysts

   Seoin Back, Junwoong Yoon, Nianhan Tian, and 3 more authors

   The Journal of Physical Chemistry Letters, Jul 2019

   DOI