Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-10318
For citation please use:
Main Title: Quantum-Chemical Insights from Interpretable Atomistic Neural Networks
Author(s): Schütt, Kristof T.
Gastegger, Michael
Tkatchenko, Alexandre
Müller, Klaus-Robert
Type: Book Part
Language Code: en
Abstract: With the rise of deep neural networks for quantum chemistry applications, there is a pressing need for architectures that, beyond delivering accurate predictions of chemical properties, are readily interpretable by researchers. Here, we describe interpretation techniques for atomistic neural networks on the example of Behler–Parrinello networks as well as the end-to-end model SchNet. Both models obtain predictions of chemical properties by aggregating atom-wise contributions. These latent variables can serve as local explanations of a prediction and are obtained during training without additional cost. Due to their correspondence to well-known chemical concepts such as atomic energies and partial charges, these atom-wise explanations enable insights not only about the model but more importantly about the underlying quantum-chemical regularities. We generalize from atomistic explanations to 3d space, thus obtaining spatially resolved visualizations which further improve interpretability. Finally, we analyze learned embeddings of chemical elements that exhibit a partial ordering that resembles the order of the periodic table. As the examined neural networks show excellent agreement with chemical knowledge, the presented techniques open up new venues for data-driven research in chemistry, physics and materials science.
URI: https://depositonce.tu-berlin.de/handle/11303/11437
http://dx.doi.org/10.14279/depositonce-10318
Issue Date: 10-Sep-2019
Date Available: 19-Jun-2020
DDC Class: 541 Physikalische Chemie
006 Spezielle Computerverfahren
Subject(s): machine learning
chemistry
explanation
Sponsor/Funder: EC/H2020/792572/EU/Machine Learning for Catalytic Carbon Dioxide Activation/MachineCat
BMBF, 01IS14013A, Verbundprojekt: BBDC - Berliner Kompetenzzentrum für Big Data
BMBF, 01IS18037A, Verbundprojekt BIFOLD-BZML: Berlin Institute for the Foundations of Learning and Data
EC/H2020/725291/EU/Beyond Static Molecules: Modeling Quantum Fluctuations in Complex Molecular Environments/BeStMo
License: http://rightsstatements.org/vocab/InC/1.0/
Book Title: Explainable AI: Interpreting, Explaining and Visualizing Deep Learning
Editor: Samek, Wojciech
Montavon, Grégoire
Vedaldi, Andrea
Hansen, Lars Kai
Müller, Klaus-Robert
Publisher: Springer
Publisher Place: Cham
Publisher DOI: 10.1007/978-3-030-28954-6_17
Page Start: 311
Page End: 330
Series: Lecture Notes in Computer Science
Series Number: 11700
EISSN: 1611-3349
ISBN: 978-3-030-28954-6
978-3-030-28953-9
ISSN: 0302-9743
Appears in Collections:FG Maschinelles Lernen » Publications

Files in This Item:
schuett_etal_2019.pdf

Accepted manuscript

Format: Adobe PDF | Size: 4.5 MB
DownloadShow Preview
Thumbnail

Item Export Bar

Items in DepositOnce are protected by copyright, with all rights reserved, unless otherwise indicated.