dc.contributor.author
Miller, Benjamin Kurt
dc.date.accessioned
2020-04-24T08:50:00Z
dc.date.available
2020-04-24T08:50:00Z
dc.identifier.uri
https://refubium.fu-berlin.de/handle/fub188/27140
dc.identifier.uri
http://dx.doi.org/10.17169/refubium-26900
dc.description.abstract
e3nn is an artificial neural network which operates on atomic coordinates and achieves equivariance to the special euclidean group in three dimensions by using spherical harmonics as features. The main experiment is to benchmark the model against a standard chemical data set called QM9, on which e3nn achieves state of the art performance on three of twelve regression targets. Along with empirical results, this thesis presents theoretical argumentation for why e3nn outperforms its closest relatives, SchNet and Cormorant, on some regression targets. Significant background regarding machine learning, quantum chemistry, and the special euclidean group is also presented.
en
dc.format.extent
54 Seiten
dc.rights.uri
http://www.fu-berlin.de/sites/refubium/rechtliches/Nutzungsbedingungen
dc.subject
deep learning
en
dc.subject
neural network
en
dc.subject
quantum chemistry
en
dc.subject
machine learning
en
dc.subject
equivariance
en
dc.subject.ddc
000 Computer science, information, and general works::000 Computer Science, knowledge, systems::004 Data processing and Computer science
dc.subject.ddc
600 Technology, Medicine, Applied sciences::600 Technology::600 Technology, Medicine, Applied sciences
dc.subject.ddc
500 Natural sciences and mathematics::540 Chemistry and allied sciences::541 Physical and theoretical chemistry
dc.title
SE(3) Equivariant Neural Networks for Regression on Molecular Properties
dc.identifier.urn
urn:nbn:de:kobv:188-refubium-27140-3
dc.title.subtitle
The QM9 Benchmark
refubium.affiliation
Mathematik und Informatik
refubium.affiliation.other
Artificial Intelligence for the Sciences
refubium.affiliation.other
Institut für Mathematik / Arbeitsgruppe Computational Molecular Biology
refubium.resourceType.isindependentpub
yes
dcterms.accessRights.dnb
free
dcterms.accessRights.openaire
open access