Chapter 2: Outliers in Machine Learning and Hysteresis
Chapter 2: Outliers in Machine Learning and Hysteresis
Embedded Files
When I extended the machine learning approach to the prediction of n-hexane at 298 K, we discovered outliers in the predictions. We believed that these outliers were not because the ML model was inaccurate but because the ML model discovered some physics.
By calculating isotherms, we found hysteresis loops for the outlier structures. The adsorption and desorption branches do not match.
A paper that includes this research is in review.
K. Shi, Z. Li, D. Anstine, C. Colina, D. Sholl, J. I. Siepmann, and R. Q. Snurr. Two-dimensional Energy Histograms as Features for Machine Learning to Predict Adsorption in Diverse Nanoporous Materials, J. Chem. Theory Comput. 2023, 19, 14, 4568–4583 (LINK)
Page updated
Google Sites
Report abuse