Constant Size Molecular Descriptors For Use With Machine Learning

Collins, Christopher R. and Gordon, Geoffrey J. and von Lilienfeld, O. Anatole and Yaron, David J. . (2018) Constant Size Molecular Descriptors For Use With Machine Learning. Journal of Chemical Physics, 148. p. 241718.

Full text not available from this repository.

Official URL: https://edoc.unibas.ch/68689/

Downloads: Statistics Overview


A set of molecular descriptors whose length is independent of molecular size is developed for machine learning models that target thermodynamic and electronic properties of molecules. These features are evaluated by monitoring performance of kernel ridge regression models on well-studied data sets of small organic molecules. The features include connectivity counts, which require only the bonding pattern of the molecule, and encoded distances, which summarize distances between both bonded and non-bonded atoms and so require the full molecular geometry. In addition to having constant size, these features summarize information regarding the local environment of atoms and bonds, such that models can take advantage of similarities resulting from the presence of similar chemical fragments across molecules. Combining these two types of features leads to models whose performance is comparable to or better than the current state of the art. The features introduced here have the advantage of leading to models that may be trained on smaller molecules and then used successfully on larger molecules.
Faculties and Departments:05 Faculty of Science > Departement Chemie
05 Faculty of Science > Departement Chemie > Former Organization Units Chemistry > Physikalische Chemie (Lilienfeld)
UniBasel Contributors:von Lilienfeld, Anatole
Item Type:Article, refereed
Article Subtype:Research Article
Publisher:AIP Publishing
Note:Publication type according to Uni Basel Research Database: Journal article
Related URLs:
Identification Number:
Last Modified:21 Aug 2019 14:46
Deposited On:21 Aug 2019 14:31

Repository Staff Only: item control page