Premium
Active Learning A Neural Network Model for Gold Clusters & Bulk From Sparse First Principles Training Data
Author(s) -
Loeffler Troy D.,
Manna Sukriti,
Patra Tarak K.,
Chan Henry,
Narayanan Badri,
Sankaranarayanan Subramanian
Publication year - 2020
Publication title -
chemcatchem
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.497
H-Index - 106
eISSN - 1867-3899
pISSN - 1867-3880
DOI - 10.1002/cctc.202001468
Subject(s) - artificial neural network , workflow , computer science , fidelity , training (meteorology) , feature (linguistics) , cover (algebra) , on the fly , artificial intelligence , training set , monte carlo method , machine learning , physics , mathematics , engineering , mechanical engineering , telecommunications , linguistics , philosophy , statistics , database , meteorology , operating system
The Cover Feature shows an active learning strategy to train a neural network model to describe the diverse geometries of gold catalytic clusters. In their Full Paper, T. Loeffler, S. Manna et al. develop a method to enable on‐the‐fly training of a neural network (NN) from sparse high‐fidelity training data obtained from first principles calculations. Their workflow is initiated with a sparse training dataset (1 to 5 data points) and is updated on‐the‐fly via a Nested Ensemble Monte Carlo scheme that iteratively queries the energy landscape in regions of failure and updates the training pool to improve the network performance. Their strategy allows development of an NN model that accurately captures the diverse size‐dependent geometries of gold clusters, as well as bulk.More information can be found in the Full Paper by T. Loeffler, S. Manna et al.