Prediction of atomization energy using graph kernel and active learning

Abstract

Data-driven prediction of molecular properties presents unique challenges to the design of machine learning methods concerning data structure/dimensionality, symmetry adaption, and confidence management. In this paper, we present a kernel-based pipeline that can learn and predict the atomization energy of molecules with high accuracy. The framework employs Gaussian process regression to perform predictions based on the similarity between molecules, which is computed using the marginalized graph kernel. To apply the marginalized graph kernel, a spatial adjacency rule is first applied to convert molecules into graphs whose vertices and edges are labeled by elements and interatomic distances, respectively. We then derive formulas for the efficient evaluation of the kernel. Specific functional components for the marginalized graph kernel are proposed, while the effect of the associated hyperparameters on accuracy and predictive confidence are examined. We show that the graph kernel is particularly suitable for predicting extensive properties because its convolutional structure coincides with that of the covariance formula between sums of random variables.Using an active learning procedure, we demonstrate that the proposed method can achieve a mean absolute error of 0.62 ± 0.01 kcal/mol using as few as 2000 training samples on the QM7 data set.

A Comparative Study of Marginalized Graph Kernel and Message Passing Neural Network

“…28 More details can be found in references. 15,[25][26][27]…”

“…25 Using GraphDot, Tang and de Jong introduced an MGK for molecular atomization energy prediction using the QM7 data sets. 26 Xiang et al developed normalized marginalized graph kernels (nMGK) for molecules and constructed accurate prediction models for various thermodynamic and transport properties of pure substances. 27 In this paper, we aim to benchmark the marginalized graph kernels using the direct message passing neural network (D-MPNN) 13 as a strong baseline.…”

ML4Chem: A Machine Learning Package for Chemistry and Materials Science

“…Also, the data sets have to be designed with enough diversity to capture meaningful underlying structures from input data. Active learning protocols [43,44] are known to help in assuring this diversity. After the training procedure, and assessment of the predictive power of models by cross-validation [45] or other model validation techniques, their parameters can be stored to perform inference in unknown data.…”

Designing compact training sets for data-driven molecular property prediction through optimal exploitation and exploration

“…This ultimately allows for building a relatively small but high-quality training set (relative to the design space) to save resources on data acquisition [11,12,13,14]. For molecular property prediction, active learning has been applied to achieve better accuracy than random selection with the same training set size [15,16].…”