Posts by Collection

portfolio

Portfolio item number 1

Short description of portfolio item number 1

Portfolio item number 2

Short description of portfolio item number 2

publications

FINETUNA: fine-tuning accelerated molecular simulations

Published in Machine Learning: Science and Technology, 2022

We present an online active learning framework for accelerating the simulation of atomic systems efficiently and accurately by incorporating prior physical information learned by large-scale pre-trained graph neural network models from the Open Catalyst Project.

Recommended citation: Joseph Musielewicz*, Xiaoxiao Wang*, Tian Tian, and Zachary Ulissi "FINETUNA: fine-tuning accelerated molecular simulations" Mach. Learn.: Sci. Technol. 3 03LT01 (2022) https://doi.org/10.1088/2632-2153/ac8fe0

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

Published in Machine Learning: Science and Technology, 2022

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

Recommended citation: Yuge Hu, Joseph Musielewicz, Zachary Ulissi, and Andrew J. Medford "Robust and scalable uncertainty estimation with conformal prediction for machine-learned interatomic potentials." Mach. Learn.: Sci. Technol. 3 045028 (2022). https://doi.org/10.1088/2632-2153/aca7b1

Predictive Uncertainty Quantification for Graph Neural Network Driven Relaxed Energy Calculations

Published in Neurips 2023 AI4Science Workshop, 2023

Popular uncertainty quantification techniques are ill-suited to graph models on relaxed energy tasks for material discovery, we show distribution-free techniques allow us to benchmark and develop novel improvements for uncertainty methods.

Recommended citation: Joseph Musielewicz, Janice Lan, and Matt Uyttendaele "Predictive Uncertainty Quantification for Graph Neural Network Driven Relaxed Energy Calculations" NeurIPS 2023 AI for Science Workshop https://openreview.net/pdf?id=rdgB5BqWCw

Generalization of Graph-Based Active Learning Relaxation Strategies Across Materials

Published in Machine Learning: Science and Technology, 2024

In this study, we present a comprehensive investigation of the broader application of Finetuna, an active learning framework to accelerate structural relaxation in DFT with prior information from Open Catalyst Project pretrained graph neural networks. We explore the challenges associated with out-of-domain systems: alcohol (C>2) on metal surfaces as larger adsorbates, metal-oxides with spin polarization, and three-dimensional (3D) structures like zeolites and metal-organic-frameworks.

Recommended citation: Joseph Musielewicz*, Xiaoxiao Wang*, Richard Tran, Sudheesh Kumar Ethirajan, Xiaoyan Fu, Hilda Mera, John R. Kitchin, Rachel C. Kurchin, and Zachary W. Ulissi "Generalization of Graph-Based Active Learning Relaxation Strategies Across Materials" arXiv preprint arXiv:2311.01987 (2023). https://dx.doi.org/10.1088/2632-2153/ad37f0

Improved Uncertainty Estimation of Graph Neural Network Potentials Using Engineered Latent Space Distances

Published in arXiv, 2024

We show that conformal prediction methods, with our novel latent space distance improvements, are the most well-calibrated and efficient approach for uncertainty quantification of relaxed energy calculations

Recommended citation: Joseph Musielewicz, Janice Lan, Matt Uyttendaele, and John R. Kitchin "Improved Uncertainty Estimation of Graph Neural Network Potentials Using Engineered Latent Space Distances"; preprint arXiv:2407.10844 (2024). https://arxiv.org/abs/2407.10844

Accessing Numerical Energy Hessians with Graph Neural Network Potentials and Their Application in Heterogeneous Catalysis

Published in arXiv, 2024

We demonstrate that off-the-shelf pretrained Open Catalyst Project (OCP) machine learned potentials (MLPs) determine the Hessian with great success (58 cm−1 mean absolute error (MAE)) for intermediates adsorbed to heterogeneous catalyst surfaces. The top performing model, with a simple offset correction, gives good estimations of the vibrational entropy contribution to the Gibbs free energy with an MAE of 0.042 eV at 300 K.

Recommended citation: Joseph Musielewicz*, Brook Wander*, Raffaele Cheula, and John R. Kitchin "Accessing Numerical Energy Hessians with Graph Neural Network Potentials and Their Application in Heterogeneous Catalysis"; preprint arXiv:2410.01650 (2024). https://arxiv.org/abs/2410.01650

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015