Chemical diversity in molecular orbital energy predictions with kernel ridge regression

Abstract

Instant machine learning predictions of molecular properties are desirable for materials design, but the predictive power of the methodology is mainly tested on well-known benchmark datasets. Here, we investigate the performance of machine learning with kernel ridge regression (KRR) for the prediction of molecular orbital energies on three large datasets: the standard QM9 small organic molecules set, amino acid and dipeptide conformers, and organic crystal-forming molecules extracted from the Cambridge Structural Database. We focus on prediction of highest occupied molecular orbital (HOMO) energies, computed at density-functional level of theory. Two different representations that encode molecular structure are compared: the Coulomb matrix (CM) and the many-body tensor representation (MBTR). We find that KRR performance depends significantly on the chemistry of the underlying dataset and that the MBTR is superior to the CM, predicting HOMO energies with a mean absolute error as low as 0.09 eV. To demonstrate the power of our machine learning method, we apply our model to structures of 10k previously unseen molecules. We gain instant energy predictions that allow us to identify interesting molecules for future applications.

Data‐Driven Materials Science: Status, Challenges, and Perspectives

“…[17] Supervised learning applies in situations where a machine learning model is trained on input-output pairs from a real process to produce optimal outputs for unseen inputs. Typical applications are predictions of physical properties (like formation energies [200][201][202] or molecular properties [203][204][205][206][207] ) given the input features of a material or process (e.g., geometry, physical properties, external conditions).…”

Machine Learning for Understanding the Relationship between the Charge Transport Mobility and Electronic Energy Levels for n‐Type Organic Field‐Effect Transistors

“…[33][34][35] Our study indicated that this machine-learning strategies may provide OFET researchers supporting details to fine-tune the electronic structure and thus the charge transport property of the n-type organic materials. [33][34][35] Our study indicated that this machine-learning strategies may provide OFET researchers supporting details to fine-tune the electronic structure and thus the charge transport property of the n-type organic materials.…”

DScribe: Library of descriptors for machine learning in materials science

“…dictions for solids [5][6][7][8], accelerated molecular property prediction [9][10][11], creation of new force-fields based on quantum mechanical training data [12][13][14][15][16][17][18][19], search for catalytically active sites in nanoclusters [20][21][22][23][24] and efficient optimization of complex structures [25][26][27]. Atomistic machine learning establishes a relationship between the atomic structure of a system and its properties.…”