[PDF] Prediction Errors of Molecular Machine Learning Models Lower than Hybrid DFT Error. | Semantic Scholar (2024)

This work investigates 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.

This work evaluates a feed-forward neural network (FNN) model's prediction performance over five feature selection methods and nine ground-state properties from a public data set composed of ∼130k organic molecules and investigates the impact of choosing free-coordinate descriptors based on the Simplified Molecular Input Line Entry System (SMILES) representation.

This work presents a comprehensive comparison of machine learning models and several molecular featurization methods - sum over bonds, custom descriptors, Coulomb matrices, Bag of Bonds, and fingerprints - and concludes that the best featurizing was sum over bond counting, and the best model was kernel ridge regression.

DFT-level accuracy of molecular energy prediction can be achieved using force-field optimized geometries and atomic vector representations learned from deep tensor neural network, and integrated molecular modeling and machine learning would be a promising approach to develop more powerful computational tools for molecular conformation analysis.

A transferable and molecule size‐independent machine learning model bonds (B), angles (A), nonbonded (N) interactions, and dihedrals (D) neural network (BAND NN) based on a chemically intuitive representation inspired by molecular mechanics force fields is presented.

This work presents a Free energy Machine Learning model applicable throughout chemical compound space and based on a representation that employs Boltzmann averages to account for an approximated sampling of configurational space and showcases its usefulness through analysis of solvation free energies.

Two different classes of molecular representations for use in machine learning of thermodynamic and electronic properties are studied, including the Coulomb matrix and Bag of Bonds, and Encoded Bonds, which encode such lists into a feature vector whose length is independent of molecular size.

A multilevel attention neural network is proposed, named DeepMoleNet, to enable chemical interpretable insights being fused into multitask learning through weighting contributions from various atoms and taking the atom-centered symmetry functions (ACSFs) as the teacher descriptor.

The hierarchical pathway from generating the dataset of reference calculations to the construction of the machine learning model, and the validation of the physics generated by the model are discussed, and empirical evidence that a higher level of theory generates a smoother PES is provided.

A series of revised autocorrelation functions (RACs) that encode relationships of the heuristic atomic properties (e.g., size, connectivity, and electronegativity) on a molecular graph are introduced to make these RACs amenable to inorganic chemistry.

...

...

Numerical evidence is presented that ML model predictions for all properties can reach an approximation error to DFT which is on par with chemical accuracy.

This work introduces a hierarchy of representations which meet uniqueness and target similarity criteria, and relies on interatomic many body expansions, as implemented in universal force-fields, including Bonding, Angular (BA), and higher order terms to systematically control target similarity.

A number of established machine learning techniques are outlined and the influence of the molecular representation on the methods performance is investigated, finding the best methods achieve prediction errors of 3 kcal/mol for the atomization energies of a wide variety of molecules.

The transferability of the approach is demonstrated, using semiempirical quantum chemistry and machine learning models trained on 1 and 10% of 134k organic molecules, to reproduce enthalpies of all remaining molecules at density functional theory level of accuracy.

Property-independent kernels for machine learning models of arbitrarily many molecular properties enable the instantaneous generation of machine learning Models which can be systematically improved through the addition of more data.

The ML-OM2 results show improved average accuracy and a much reduced error range compared with those of standard OM2 results, with mean absolute errors in atomization enthalpies dropping from 6.3 to 1.7 kcal/mol.

A systematic hierarchy of efficient empirical methods to estimate atomization and total energies of molecules and is achieved by a vectorized representation of molecules (so-called Bag of Bonds model) that exhibits strong nonlocality in chemical space.

A set of molecular descriptors whose length is independent of molecular size is developed for machine learning models that target thermodynamic and electronic properties of molecules, which have the advantage of leading to models that may be trained on smaller molecules and then used successfully on larger molecules.

A fingerprint representation of molecules based on a Fourier series of atomic radial distribution functions that requires no preconceived knowledge about chemical bonding, topology, or electronic orbitals is introduced, suggesting its usefulness for machine learning models of molecular properties trained across chemical compound space.

A machine learning model is introduced to predict atomization energies of a diverse set of organic molecules, based on nuclear charges and atomic positions only, and applicability is demonstrated for the prediction of molecular atomization potential energy curves.

...

...