Supporting information for a multifidelity neural network formulation for molecular potential energy surfaces

This is a supplementary information for our paper titled "<em>Multifidelity neural network formulations for prediction of quantum chemistry potential energy surfaces</em>" Supplemental information includes two data files corresponding to the complete sets of low and high fidelity training data used in numerical experiments. Format is JavaScript Object Notation (JSON). 1. low_fidelity_training_data.json contains 74000 records 2. high_fidelity_training_data.json contains 36988 records Each record consists of a numerical id ("id"), (x,y,z) position tuples ("geometry") for C5H5 ordered as 5 carbon atoms followed by 5 hydrogen atoms, and corresponding potential energy ("energy"). Source: normal mode sampling around 2 wells, 1 transition state, and a set of IRCs as depicted in Figure 2. Usage: subsets of this data were used as needed to define different data amounts and different subset randomizations in Figures 5 through 8.