English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Machine learning interatomic potentials (MLIPs) are promising surrogates forquantum mechanics evaluations in ab-initio molecular dynamics simulations dueto their ability to reproduce the energy and force landscape within chemicalaccuracy at four orders of magnitude less cost. While developing uncertaintyquantification (UQ) tools for MLIPs is critical to build production MLIPdatasets using active learning, only limited progress has been made and themost robust method, ensembling, still shows low correlation between high errorand high uncertainty predictions. Here we develop a rigorous method rooted instatistics for determining an error cutoff that distinguishes regions of highand low UQ performance. The statistical cutoff illuminates that a main cause ofthe poor UQ performance is due to the machine learning model already describingthe entire dataset and not having any datapoints with error greater than thestatistical error distribution. Second, we extend the statistical analysis tocreate an interpretable connection between the error and uncertaintydistributions to predict an uncertainty cutoff separating high and low errors.We showcase the statistical cutoff in active learning benchmarks on twodatasets of varying chemical complexity for three common UQ methods:ensembling, sparse Gaussian processes, and latent distance metrics and comparethem to the true error and random sampling, showing that the statistical cutoffis generalizable to a variety of different UQ methods and protocols andperforms similarly to using the true error. Importantly, we conclude thatutilizing this uncertainty cutoff enables using significantly lower costuncertainty quantification tools such as sparse gaussian processes and latentdistances compared to ensembling approaches for generating MLIP datasets at afraction of the cost.

Statistical methods for resolving poor uncertainty quantification in machine learning interatomic potentials