Deep Learning Spectroscopy: Neural Networks for Molecular Excitation Spectra
Top Cited Papers
Open Access
- 29 January 2019
- journal article
- research article
- Published by Wiley in Advanced Science
- Vol. 6 (9) , 1801367
Abstract
Deep learning methods for the prediction of molecular excitation spectra are presented. For the example of the electronic density of states of 132k organic molecules, three different neural network architectures: multilayer perceptron (MLP), convolutional neural network (CNN), and deep tensor neural network (DTNN) are trained and assessed. The inputs for the neural networks are the coordinates and charges of the constituent atoms of each molecule. Already, the MLP is able to learn spectra, but the root mean square error (RMSE) is still as high as 0.3 eV. The learning quality improves significantly for the CNN (RMSE = 0.23 eV) and reaches its best performance for the DTNN (RMSE = 0.19 eV). Both CNN and DTNN capture even small nuances in the spectral shape. In a showcase application of this method, the structures of 10k previously unseen organic molecules are scanned and instant spectra predictions are obtained to identify molecules for potential applications.Keywords
Funding Information
- Horizon 2020 Framework Programme (676580)
- Research Council of Finland (284621, 305632)
This publication has 49 references indexed in Scilit:
- Machine Learning for Quantum Mechanical Properties of Atoms in MoleculesThe Journal of Physical Chemistry Letters, 2015
- Big Data Meets Quantum Chemistry Approximations: The Δ-Machine Learning ApproachJournal of Chemical Theory and Computation, 2015
- Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical ForcesPhysical Review Letters, 2015
- Quantum chemistry structures and properties of 134 kilo moleculesScientific Data, 2014
- Machine learning of molecular electronic properties in chemical compound spaceNew Journal of Physics, 2013
- Assessment and Validation of Machine Learning Methods for Predicting Molecular Atomization EnergiesJournal of Chemical Theory and Computation, 2013
- New developments in evolutionary structure prediction algorithm USPEXComputer Physics Communications, 2013
- Compressive sensing as a paradigm for building physics modelsPhysical Review B, 2013
- Fast and Accurate Modeling of Molecular Atomization Energies with Machine LearningPhysical Review Letters, 2012
- Gaussian Approximation Potentials: The Accuracy of Quantum Mechanics, without the ElectronsPhysical Review Letters, 2010