Learning invariant representations of molecules for atomization energy prediction

Grègoire Montavon; Katja Hansen; Siamac Fazli; Matthias Rupp; Franziska Biegler; Andreas Ziehe; Alexandre Tkatchenko; O. Anatole Von Lilienfeld; Klaus Robert Müller

Learning invariant representations of molecules for atomization energy prediction

Grègoire Montavon, Katja Hansen, Siamac Fazli, Matthias Rupp, Franziska Biegler, Andreas Ziehe, Alexandre Tkatchenko, O. Anatole Von Lilienfeld, Klaus Robert Müller

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

108 Citations (Scopus)

Abstract

The accurate prediction of molecular energetics in chemical compound space is a crucial ingredient for rational compound design. The inherently graph-like, non-vectorial nature of molecular data gives rise to a unique and difficult machine learning problem. In this paper, we adopt a learning-from-scratch approach where quantum-mechanical molecular energies are predicted directly from the raw molecular geometry. The study suggests a benefit from setting flexible priors and enforcing invariance stochastically rather than structurally. Our results improve the state-of-the-art by a factor of almost three, bringing statistical methods one step closer to chemical accuracy.

Original language	English
Title of host publication	Advances in Neural Information Processing Systems 25
Subtitle of host publication	26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012
Pages	440-448
Number of pages	9
Publication status	Published - Dec 1 2012
Externally published	Yes
Event	26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012 - Lake Tahoe, NV, United States Duration: Dec 3 2012 → Dec 6 2012

Publication series

Name	Advances in Neural Information Processing Systems
Volume	1
ISSN (Print)	1049-5258

Conference

Conference	26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012
Country/Territory	United States
City	Lake Tahoe, NV
Period	12/3/12 → 12/6/12

ASJC Scopus subject areas

Computer Networks and Communications
Information Systems
Signal Processing

Cite this

Montavon, G., Hansen, K., Fazli, S., Rupp, M., Biegler, F., Ziehe, A., Tkatchenko, A., Von Lilienfeld, O. A., & Müller, K. R. (2012). Learning invariant representations of molecules for atomization energy prediction. In Advances in Neural Information Processing Systems 25: 26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012 (pp. 440-448). (Advances in Neural Information Processing Systems; Vol. 1).

Learning invariant representations of molecules for atomization energy prediction. / Montavon, Grègoire; Hansen, Katja; Fazli, Siamac et al.
Advances in Neural Information Processing Systems 25: 26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012. 2012. p. 440-448 (Advances in Neural Information Processing Systems; Vol. 1).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Montavon, G, Hansen, K, Fazli, S, Rupp, M, Biegler, F, Ziehe, A, Tkatchenko, A, Von Lilienfeld, OA & Müller, KR 2012, Learning invariant representations of molecules for atomization energy prediction. in Advances in Neural Information Processing Systems 25: 26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012. Advances in Neural Information Processing Systems, vol. 1, pp. 440-448, 26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012, Lake Tahoe, NV, United States, 12/3/12.

@inproceedings{dd9bf8a8c9fa4d0984e46319a4182aca,

title = "Learning invariant representations of molecules for atomization energy prediction",

abstract = "The accurate prediction of molecular energetics in chemical compound space is a crucial ingredient for rational compound design. The inherently graph-like, non-vectorial nature of molecular data gives rise to a unique and difficult machine learning problem. In this paper, we adopt a learning-from-scratch approach where quantum-mechanical molecular energies are predicted directly from the raw molecular geometry. The study suggests a benefit from setting flexible priors and enforcing invariance stochastically rather than structurally. Our results improve the state-of-the-art by a factor of almost three, bringing statistical methods one step closer to chemical accuracy.",

author = "Gr{\`e}goire Montavon and Katja Hansen and Siamac Fazli and Matthias Rupp and Franziska Biegler and Andreas Ziehe and Alexandre Tkatchenko and {Von Lilienfeld}, {O. Anatole} and M{\"u}ller, {Klaus Robert}",

year = "2012",

month = dec,

day = "1",

language = "English",

isbn = "9781627480031",

series = "Advances in Neural Information Processing Systems",

pages = "440--448",

booktitle = "Advances in Neural Information Processing Systems 25",

note = "26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012 ; Conference date: 03-12-2012 Through 06-12-2012",

}

TY - GEN

T1 - Learning invariant representations of molecules for atomization energy prediction

AU - Montavon, Grègoire

AU - Hansen, Katja

AU - Fazli, Siamac

AU - Rupp, Matthias

AU - Biegler, Franziska

AU - Ziehe, Andreas

AU - Tkatchenko, Alexandre

AU - Von Lilienfeld, O. Anatole

AU - Müller, Klaus Robert

PY - 2012/12/1

Y1 - 2012/12/1

N2 - The accurate prediction of molecular energetics in chemical compound space is a crucial ingredient for rational compound design. The inherently graph-like, non-vectorial nature of molecular data gives rise to a unique and difficult machine learning problem. In this paper, we adopt a learning-from-scratch approach where quantum-mechanical molecular energies are predicted directly from the raw molecular geometry. The study suggests a benefit from setting flexible priors and enforcing invariance stochastically rather than structurally. Our results improve the state-of-the-art by a factor of almost three, bringing statistical methods one step closer to chemical accuracy.

AB - The accurate prediction of molecular energetics in chemical compound space is a crucial ingredient for rational compound design. The inherently graph-like, non-vectorial nature of molecular data gives rise to a unique and difficult machine learning problem. In this paper, we adopt a learning-from-scratch approach where quantum-mechanical molecular energies are predicted directly from the raw molecular geometry. The study suggests a benefit from setting flexible priors and enforcing invariance stochastically rather than structurally. Our results improve the state-of-the-art by a factor of almost three, bringing statistical methods one step closer to chemical accuracy.

UR - http://www.scopus.com/inward/record.url?scp=84877770101&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84877770101&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84877770101

SN - 9781627480031

T3 - Advances in Neural Information Processing Systems

SP - 440

EP - 448

BT - Advances in Neural Information Processing Systems 25

T2 - 26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012

Y2 - 3 December 2012 through 6 December 2012

ER -

Learning invariant representations of molecules for atomization energy prediction

Abstract

Publication series

Conference

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this