molecular design; sparse grid combination technique; machine learning
Harbrecht Helmut, Multerer Michael (2021), A fast direct solver for nonlocal operators in wavelet coordinates, in Journal of Computational Physics
, 428, 110056-110056.
von Lilienfeld O. Anatole, Burke Kieron (2020), Retrospective on a decade of machine learning for chemical discovery, in Nature Communications
, 11(1), 4895-4895.
Huang Bing, von Lilienfeld O. Anatole (2020), Quantum machine learning using atom-in-molecule-based fragments selected on the fly, in Nature Chemistry
, 12(10), 945-951.
Heinen Stefan, Schwilk Max, von Rudorff Guido Falk, von Lilienfeld O Anatole (2020), Machine learning the computational cost of quantum chemistry, in Machine Learning: Science and Technology
, 1(2), 025002-025002.
Harbrecht Helmut, Zaspel Peter (2019), On the Algebraic Construction of Sparse Multilevel Approximations of Elliptic Tensor Product Problems, in Journal of Scientific Computing
, 78(2), 1272-1290.
Zaspel Peter (2019), Algorithmic Patterns for H-Matrices on Many-Core Processors, in Journal of Scientific Computing
, 78(2), 1174-1206.
Zaspel Peter, Huang Bing, Harbrecht Helmut, von Lilienfeld O. Anatole (2019), Boosting Quantum Machine Learning Models with a Multilevel Combination Technique: Pople Diagrams Revisited, in Journal of Chemical Theory and Computation
, 15(3), 1546-1559.
Zaspel Peter (2019), ENSEMBLE KALMAN FILTERS FOR RELIABILITY ESTIMATION IN PERFUSION INFERENCE, in International Journal for Uncertainty Quantification
, 9(1), 15-32.
HarbrechtHelmut, JakemanJohn Davis, ZaspelPeter, Cholesky-based experimental design for Gaussian process and kernel-based emulation and calibration, in Commun. Comput. Phys.
We propose to sample an unprecedented amounts of molecular data (166 B molecules) with a unified multilevel (sparse grids plus combination rules)/machine learning based approach. A chemically accurate and transferable property model will be generated on-the-fly, allowing for unprecedented computational efficiency. Subsequently, we will demonstrate the superior performance of this approach by its use for surrogate models within iterative optimization solvers which enable the computational design of new molecules with desired properties in real time. The objective for this effort is twofold: We would like to (a) provide the experimental chemist with a powerful tool to guide their design, synthesis, and characterization efforts, and to (b) discover and gain a better understanding about established as well as hereto unknown trends and relationships between chemical and physical properties throughout chemical space.