Searcharxiv⌕ Search

arXiv subjects

O. Anatole von Lilienfeld

Publications and source records attributed to O. Anatole von Lilienfeld.

At least 73 records · Page 4Linked to original sources
arXiv2020

Wasserstein metric for improved QML with adjacency matrix representations

We study the Wasserstein metric to measure distances between molecules represented by the atom index dependent adjacency "Coulomb" matrix, used in kernel ridge regression based supervised learning. Resulting quantum machine learning models exhibit improved training efficiency and result in smoother predictions of molecular distortions. We first demonstrate smoothness for the continuous extraction of an atom from some organic molecule. Learning curves, quantifying the decay of the atomization energy's prediction error as a function of training set size, have been obtained for tens of thousands of organic molecules drawn from the QM9 data set. In comparison to conventionally used metrics ($L_1$ and $L_2$ norm), our numerical results indicate systematic improvement in terms of learning curve off-set for random as well as sorted (by norms of row) atom indexing in Coulomb matrices. Our findings suggest that this metric corresponds to a favorable similarity measure which introduces index-invariance in any kernel based model relying on adjacency matrix representations.

physics.chem-ph↗
arXiv2020

FCHL revisited: faster and more accurate quantum machine learning

We introduce the FCHL19 representation for atomic environments in molecules or condensed-phase systems. Machine learning models based on FCHL19 are able to yield predictions of atomic forces and energies of query compounds with chemical accuracy on the scale of milliseconds. FCHL19 is a revision of our previous work [Faber et al. 2018] where the representation is discretized and the individual features are rigorously optimized using Monte Carlo optimization. Combined with a Gaussian kernel function that incorporates elemental screening, chemical accuracy is reached for energy learning on the QM7b and QM9 datasets after training for minutes and hours, respectively. The model also shows good performance for non-bonded interactions in the condensed phase for a set of water clusters with an MAE binding energy error of less than 0.1 kcal/mol/molecule after training on 3,200 samples. For force learning on the MD17 dataset, our optimized model similarly displays state-of-the-art accuracy with a regressor based on Gaussian process regression. When the revised FCHL19 representation is combined with the operator quantum machine learning regressor, forces and energies can be predicted in only a few milliseconds per atom. The model presented herein is fast and lightweight enough for use in general chemistry problems as well as molecular dynamics simulations.

physics.chem-ph↗
arXiv2019

Rapid and accurate molecular deprotonation energies from quantum alchemy

We assess the applicability of Alchemical Perturbation Density Functional Theory (APDFT) for quickly and accurately estimating deprotonation energies. We have considered all possible single and double deprotonations in one hundred small organic molecules drawn at random from QM9 [Ramakrishnan et al, JCTC 2015]. Numerical evidence is presented for 5'160 deprotonated species at both HF/def2-TZVP and CCSD/6-31G* level of theory. We show that the perturbation expansion formalism of APDFT quickly converges to reliable results: using CCSD electron densities and derivatives, regular Hartree-Fock is outperformed at second or third order for ranking all possible doubly or singly deprotonated molecules, respectively. CCSD single deprotonation energies are reproduced within 1.4 kcal/mol on average within third order APDFT. We introduce a hybrid approach were the computational cost of APDFT is reduced even further by mixing first order terms at a higher level of theory (CCSD) with higher order terms at a lower level of theory only (HF). We find that this approach reaches 2 kcal/mol accuracy in absolute deprotonation energies compared to CCSD at 2% of the computational cost of third order APDFT.

physics.chem-ph↗
arXiv2019

Machine learning the computational cost of quantum chemistry

Computational quantum mechanics based molecular and materials design campaigns consume increasingly more high-performance compute resources, making improved job scheduling efficiency desirable in order to reduce carbon footprint or wasteful spending. We introduce quantum machine learning (QML) models of the computational cost of common quantum chemistry tasks. For 2D non-linear toy systems, single point, geometry optimization, and transition state calculations the out of sample prediction error of QML models of wall times decays systematically with training set size. We present numerical evidence for a toy system containing two functions and three commonly used optimizer and for thousands of organic molecular systems including closed and open shell equilibrium structures, as well as transition states. Levels of electronic structure theory considered include B3LYP/def2-TZVP, MP2/6-311G(d), local CCSD(T)/VTZ-F12, CASSCF/VDZ-F12, and MRCISD+Q-F12/VDZ-F12. In comparison to conventional indiscriminate job treatment, QML based wall time predictions significantly improve job scheduling efficiency for all tasks after training on just thousands of molecules. Resulting reductions in CPU time overhead range from 10% to 90%.

physics.chem-ph↗
arXiv2019

Neural networks and kernel ridge regression for excited states dynamics of CH$_2$NH$_2^+$: From single-state to multi-state representations and multi-property machine learning models

Excited-state dynamics simulations are a powerful tool to investigate photo-induced reactions of molecules and materials and provide complementary information to experiments. Since the applicability of these simulation techniques is limited by the costs of the underlying electronic structure calculations, we develop and assess different machine learning models for this task. The machine learning models are trained on {\emph ab initio} calculations for excited electronic states, using the methylenimmonium cation (CH$_2$NH$_2^+$) as a model system. For the prediction of excited-state properties, multiple outputs are desirable, which is straightforward with neural networks but less explored with kernel ridge regression. We overcome this challenge for kernel ridge regression in the case of energy predictions by encoding the electronic states explicitly in the inputs, in addition to the molecular representation. We adopt this strategy also for our neural networks for comparison. Such a state encoding enables not only kernel ridge regression with multiple outputs but leads also to more accurate machine learning models for state-specific properties. An important goal for excited-state machine learning models is their use in dynamics simulations, which needs not only state-specific information but also couplings, i.e., properties involving pairs of states. Accordingly, we investigate the performance of different models for such coupling elements. Furthermore, we explore how combining all properties in a single neural network affects the accuracy. As an ultimate test for our machine learning models, we carry out excited-state dynamics simulations based on the predicted energies, forces and couplings and, thus, show the scopes and possibilities of machine learning for the treatment of electronically excited states.

physics.chem-ph↗
arXiv2019

Exploring Chemical Compound Space with Quantum-Based Machine Learning

Rational design of compounds with specific properties requires conceptual understanding and fast evaluation of molecular properties throughout chemical compound space (CCS) -- the huge set of all potentially stable molecules. Recent advances in combining quantum mechanical (QM) calculations with machine learning (ML) provide powerful tools for exploring wide swaths of CCS. We present our perspective on this exciting and quickly developing field by discussing key advances in the development and applications of QM-based ML methods to diverse compounds and properties and outlining the challenges ahead. We argue that significant progress in the exploration and understanding of CCS can be made through a systematic combination of rigorous physical theories, comprehensive synthetic datasets of microscopic and macroscopic properties, and modern ML methods that account for physical and chemical knowledge.

physics.chem-ph↗
arXiv2019

Operator quantum machine learning: Navigating the chemical space of response properties

The identification and use of structure property relationships lies at the heart of the chemical sciences. Quantum mechanics forms the basis for the unbiased virtual exploration of chemical compound space (CCS), imposing substantial compute needs if chemical accuracy is to be reached. In order to accelerate predictions of quantum properties without compromising accuracy, our lab has been developing quantum machine learning (QML) based models which can be applied throughout CCS. Here, we briefly explain, review, and discuss the recently introduced operator formalism which substantially improves the data efficiency for QML models of common response properties.

physics.chem-ph↗
arXiv2019

Atoms in molecules from alchemical perturbation density functional theory

Based on thermodynamic integration we introduce atoms in molecules (AIM) using the orbital-free framework of alchemical perturbation density functional theory (APDFT). Within APDFT, atomic energies and electron densities in molecules are arbitrary because any arbitrary reference system and integration path can be selected as long as it meets the boundary conditions. We choose the uniform electron gas as the most generic reference, and linearly scale up all nuclear charges, situated at any query molecule's atomic coordinates. Within the approximations made when calculating one-particle electron densities, this choice affords exact and unambiguous definitions of energies and electron densities of AIMs We illustrate the approach for neutral iso-electronic diatomics (CO, N$_2$, BF), various small molecules with different electronic hybridisation states of carbon (CH$_4$, C$_2$H$_6$, C$_2$H$_4$, C$_2$H$_2$, HCN), and for all the possible BN doped mutants connecting benzene to borazine (C$_{2n}$B$_{3-n}$N$_{3-n}$H$_6$, $0 \le n \le 3$). Analysis of the numerical results obtained suggests that APDFT based AIMs enable meaningful and new interpretations of molecular energies and electron densities.

physics.chem-ph↗
arXiv2019

Alchemical perturbation density functional theory (APDFT)

We introduce an orbital free electron density functional approximation based on alchemical perturbation theory. Given convergent perturbations of a suitable reference system, the accuracy of popular self-consistent Kohn-Sham density functional estimates of properties of new molecules can be systematically surpassed---at negligible cost. The associated energy functional is an approximation to the integrated energy derivative, requiring only perturbed reference electron densities: No self-consistent field equations are necessary to estimate energies and electron densities. Electronic ground state properties considered include covalent bonding potentials, atomic forces, as well as dipole and quadropole moments.

physics.chem-ph↗
arXiv2018

Alchemical normal modes unify chemical space

In silico design of new molecules and materials with desirable quantum properties by high-throughput screening is a major challenge due to the high dimensionality of chemical space. To facilitate its navigation, we present a unification of coordinate and composition space in terms of alchemical normal modes (ANMs) which result from second order perturbation theory. ANMs assume a predominantly smooth nature of chemical space and form a basis in which new compounds can be expanded and identified. We showcase the use of ANMs for the energetics of the iso-electronic series of diatomics with 14 electrons, BN doped benzene derivatives (C$_{6-2x}$(BN)$_{x}$H$_6$ with $x = 0, 1, 2, 3$), predictions for over 1.8 million BN doped coronene derivatives, and genetic energy optimizations in the entire BN doped coronene space. Using Ge lattice scans as reference, the applicability ANMs across the periodic table is demonstrated for III-V and IV-IV-semiconductors Si, Sn, SiGe, SnGe, SiSn, as well as AlP, AlAs, AlSb, GaP, GaAs, GaSb, InP, InAs, and InSb. Analysis of our results indicates simple qualitative structure property rules for estimating energetic rankings among isomers. Useful quantitative estimates can also be obtained when few atoms are changed to neighboring or lower lying elements in the periodic table. The quality of the predictions often increases with the symmetry of system chosen as reference due to cancellation of odd order terms. Rooted in perturbation theory the ANM approach promises to generally enable unbiased compound exploration campaigns at reduced computational cost.

physics.chem-ph↗
arXiv2018

Operators in Machine Learning: Response Properties in Chemical Space

The role of response operators is well established in quantum mechanics. We investigate their use for universal quantum machine learning models of response properties in molecules. After introducing a theoretical basis, we present and discuss numerical evidence based on measuring the potential energy's response with respect to atomic displacement and to electric fields. Prediction errors for corresponding properties, atomic forces and dipole moments, improve in a systematic fashion with training set size and reach high accuracy for small training sets. Prediction of normal modes and IR-spectra of some small molecules demonstrates the usefulness of this approach for chemistry.

physics.chem-ph↗
arXiv2018

Boosting quantum machine learning models with multi-level combination technique: Pople diagrams revisited

Inspired by Pople diagrams popular in quantum chemistry, we introduce a hierarchical scheme, based on the multi-level combination (C) technique, to combine various levels of approximations made when calculating molecular energies within quantum chemistry. When combined with quantum machine learning (QML) models, the resulting CQML model is a generalized unified recursive kernel ridge regression which exploits correlations implicitly encoded in training data comprised of multiple levels in multiple dimensions. Here, we have investigated up to three dimensions: Chemical space, basis set, and electron correlation treatment. Numerical results have been obtained for atomization energies of a set of $\sim$7'000 organic molecules with up to 7 atoms (not counting hydrogens) containing CHONFClS, as well as for $\sim$6'000 constitutional isomers of C$_7$H$_{10}$O$_2$. CQML learning curves for atomization energies suggest a dramatic reduction in necessary training samples calculated with the most accurate and costly method. In order to generate milli-second estimates of CCSD(T)/cc-pvdz atomization energies with prediction errors reaching chemical accuracy ($\sim$1 kcal/mol), the CQML model requires only $\sim$100 training instances at CCSD(T)/cc-pvdz level, rather than thousands within conventional QML, while more training molecules are required at lower levels. Our results suggest a possibly favourable trade-off between various hierarchical approximations whose computational cost scales differently with electron number.

physics.chem-ph↗
arXiv2018

The fundamentals of quantum machine learning

Within the past few years, we have witnessed the rising of quantum machine learning (QML) models which infer electronic properties of molecules and materials, rather than solving approximations to the electronic Schrodinger equation. The increasing availability of large quantum mechanics reference data sets have enabled these developments. We review the basic theories and key ingredients of popular QML models such as choice of regressor, data of varying trustworthiness, the role of the representation, and the effect of training set selection. Throughout we emphasize the indispensable role of learning curves when it comes to the comparative assessment of different QML models.

physics.chem-ph↗
arXiv2018

Al$_x$Ga$_{1-x}$As crystals with direct 2 eV band gaps from computational alchemy

We use alchemical first order derivatives for the rapid yet robust prediction of band structures. The power of the approach is demonstrated for the design challenge of finding Al$_x$Ga$_{1-x}$As semiconductor alloys with large direct band gap using computational alchemy within a genetic algorithm. Dozens of crystal polymorphs are identified for $x>\frac{2}{3}$ with direct band gaps larger than 2\:eV according to HSE approximated density functional theory. Based on a single generalized gradient approximated density functional theory band structure calculation of pure GaAs we observe convergence after visiting only $\sim$800 crystal candidates. The general applicability of alchemical gradients is demonstrated for band structure estimates in III-V and IV-IV crystals as well as for H$_2$ uptake in Sr and Ca-alanate crystals.

physics.chem-ph↗
arXiv2018

Non-covalent interactions across organic and biological subsets of chemical space: Physics-based potentials parametrized from machine learning

Classical intermolecular potentials typically require an extensive parametrization procedure for any new compound considered. To do away with prior parametrization, we propose a combination of physics-based potentials with machine learning (ML), coined IPML, which is transferable across small neutral organic and biologically-relevant molecules. ML models provide on-the-fly predictions for environment-dependent local atomic properties: electrostatic multipole coefficients (significant error reduction compared to previously reported), the population and decay rate of valence atomic densities, and polarizabilities across conformations and chemical compositions of H, C, N, and O atoms. These parameters enable accurate calculations of intermolecular contributions---electrostatics, charge penetration, repulsion, induction/polarization, and many-body dispersion. Unlike other potentials, this model is transferable in its ability to handle new molecules and conformations without explicit prior parametrization: All local atomic properties are predicted from ML, leaving only eight global parameters---optimized once and for all across compounds. We validate IPML on various gas-phase dimers at and away from equilibrium separation, where we obtain mean absolute errors between 0.4 and 0.7 kcal/mol for several chemically and conformationally diverse datasets representative of non-covalent interactions in biologically-relevant molecules. We further focus on hydrogen-bonded complexes---essential but challenging due to their directional nature---where datasets of DNA base pairs and amino acids yield an extremely encouraging 1.4 kcal/mol error. Finally, and as a first look, we consider IPML in denser systems: water clusters, supramolecular host-guest complexes, and the benzene crystal.

physics.chem-ph↗
arXiv2017

Alchemical and structural distribution based representation for improved QML

We introduce a representation of any atom in any chemical environment for the generation of efficient quantum machine learning (QML) models of common electronic ground-state properties. The representation is based on scaled distribution functions explicitly accounting for elemental and structural degrees of freedom. Resulting QML models afford very favorable learning curves for properties of out-of-sample systems including organic molecules, non-covalently bonded protein side-chains, (H$_2$O)$_{40}$-clusters, as well as diverse crystals. The elemental components help to lower the learning curves, and, through interpolation across the periodic table, even enable "alchemical extrapolation" to covalent bonding between elements not part of training, as evinced for single, double, and triple bonds among main-group elements.

physics.chem-ph↗
arXiv2017

Machine learning prediction errors better than DFT accuracy

We investigate the impact of choosing regressors and molecular representations for the construction of fast machine learning (ML) models of thirteen electronic ground-state properties of organic molecules. The performance of each regressor/representation/property combination is assessed using learning curves which report out-of-sample errors as a function of training set size with up to $\sim$117k distinct molecules. Molecular structures and properties at hybrid density functional theory (DFT) level of theory used for training and testing come from the QM9 database [Ramakrishnan et al, {\em Scientific Data} {\bf 1} 140022 (2014)] and include dipole moment, polarizability, HOMO/LUMO energies and gap, electronic spatial extent, zero point vibrational energy, enthalpies and free energies of atomization, heat capacity and the highest fundamental vibrational frequency. Various representations from the literature have been studied (Coulomb matrix, bag of bonds, BAML and ECFP4, molecular graphs (MG)), as well as newly developed distribution based variants including histograms of distances (HD), and angles (HDA/MARAD), and dihedrals (HDAD). Regressors include linear models (Bayesian ridge regression (BR) and linear regression with elastic net regularization (EN)), random forest (RF), kernel ridge regression (KRR) and two types of neural net works, graph convolutions (GC) and gated graph networks (GG). We present numerical evidence that ML model predictions deviate from DFT less than DFT deviates from experiment for all properties. Furthermore, our out-of-sample prediction errors with respect to hybrid DFT reference are on par with, or close to, chemical accuracy. Our findings suggest that ML models could be more accurate than hybrid DFT if explicitly electron correlated quantum (or experimental) data was available.

physics.chem-ph↗