Machine learning prediction errors better than DFT accuracy

@article{Faber2017MachineLP,
  title={Machine learning prediction errors better than DFT accuracy},
  author={F. Faber and Luke Hutchison and B. Huang and J. Gilmer and S. Schoenholz and G. Dahl and Oriol Vinyals and Steven M. Kearnes and P. Riley and O. A. V. Lilienfeld},
  journal={arXiv: Chemical Physics},
  year={2017}
}
  • F. Faber, Luke Hutchison, +7 authors O. A. V. Lilienfeld
  • Published 2017
  • Chemistry, Physics
  • arXiv: Chemical Physics
  • 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… CONTINUE READING
    13 Citations

    Figures and Tables from this paper

    Analyzing Learned Molecular Representations for Property Prediction
    • 119
    • PDF
    Deep Potential: a general representation of a many-body potential energy surface
    • 71
    • PDF
    Heterogeneous Molecular Graph Neural Networks for Predicting Molecule Properties
    • Zeren Shui, G. Karypis
    • Computer Science, Physics
    • 2020 IEEE International Conference on Data Mining (ICDM)
    • 2020
    • 2
    • PDF
    Combining Machine Learning and Physics to Understand Glassy Systems
    • 11
    • PDF
    Directional Message Passing for Molecular Graphs
    • 50
    • PDF
    Batched Sparse Matrix Multiplication for Accelerating Graph Convolutional Networks
    • 2
    • PDF
    Are Learned Molecular Representations Ready For Prime Time?
    • 9

    References

    Kernel Ridge Regression
    • V. Vovk
    • Mathematics, Computer Science
    • Empirical Inference
    • 2013
    • 90
    • PDF