Estimating the energy of a molecule is a long-standing challenge in chemistry. Usually, Hartree energy is obtained by complex calculations requiring intensive computational resources. This work investigates Deep Convolutional Neural Networks (CNN) to predict Hartree energy. To this end, Retrievium repository collected by Gaussian software [1] is employed as a dataset containing structures with different sizes. We represent the molecules as a square matrix containing the distance between atoms plus some atomic features. We employed the VGG model as a kind of benchmark model in CNN to perform the energy prediction, which revealed promising results in terms of Mean Absolute Error (MAE) and r2 score
Article ID: 2022G4
Month: May
Year: 2022
Address: Online
Venue: Graduate Student Symposium- Canadian Conference on Artificial Intelligence
Publisher: Canadian Artificial Intelligence Association
URL: https://caiac.pubpub.org/pub/uomoz29m