Reducing Training Data Needs with Minimal Multilevel Machine Learning (M3L)

Abstract

For many machine learning applications in science, data acquisition, not training, is the bottleneck even when avoiding experiments and relying on computation and simulation. Correspondingly, and in order to reduce cost and carbon footprint, training data efficiency is key. We introduce minimal multilevel machine learning (M3L) which optimizes training data set sizes using a loss function at multiple levels of reference data in order to minimize a combination of prediction error with overall training data acquisition costs (as measured by computational wall-times). Numerical evidence has been obtained for calculated atomization energies and electron affinities of thousands of organic molecules at various levels of theory including HF, MP2, DLPNO-CCSD(T), DFHFCABS, PNOMP2F12, and PNOCCSD(T)F12, and treating tens with basis sets TZ, cc-pVTZ, and AVTZ-F12. Our M3L benchmarks for reaching chemical accuracy in distinct chemical compound sub-spaces indicate substantial computational cost reductions by factors of $\sim$ 1.01, 1.1, 3.8, 13.8 and 25.8 when compared to heuristic sub-optimal multilevel machine learning (M2L) for the data sets QM7b, QM9${}^{\mathrm{LCCSD}(\mathrm{T})}$, EGP, QM9${}_{\mathrm{AE}}^{\mathrm{CCSD}(\mathrm{T})}$, and QM9${}_{\mathrm{EA}}^{\mathrm{CCSD}(\mathrm{T})}$, respectively. Furthermore, we use M2L to investigate the performance for 76 density functionals when used within multilevel learning and building on the following levels drawn from the hierarchy of Jacobs Ladder:~LDA, GGA, mGGA, and hybrid functionals. Within M2L and the molecules considered, mGGAs do not provide any noticeable advantage over GGAs. Among the functionals considered and in combination with LDA, the three on average top performing GGA and Hybrid levels for atomization energies on QM9 using M3L correspond respectively to PW91, KT2, B97D, and $\tau$-HCTH, B3LYP$\ast$(VWN5), TPSSH.

References in corpus (15)

• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• A Performance and Cost Assessment of Machine Learning Interatomic Potentials
• Big Data meets Quantum Chemistry Approximations: The $\Delta$-Machine Learning Approach
• Machine learning prediction errors better than DFT accuracy
• Machine Learning of Molecular Electronic Properties in Chemical Compound Space
• SchNet: A continuous-filter convolutional neural network for modeling quantum interactions
• SchNetPack: A Deep Learning Toolbox For Atomistic Systems
• FCHL revisited: faster and more accurate quantum machine learning
• Alchemical and structural distribution based representation for improved QML
• Equivariant message passing for the prediction of tensorial properties and molecular spectra
• Towards self-driving laboratories: The central role of density functional theory in the AI age
• Feature Optimization for Atomistic Machine Learning Yields A Data-Driven Construction of the Periodic Table of the Elements
• Machine learning the computational cost of quantum chemistry
• Kernel based quantum machine learning at record rate : Many-body distribution functionals as compact representations
• The FHI-aims Code: All-electron, ab initio materials simulations towards the exascale

Cited by in corpus (3)

• Improved decision making with similarity based machine learning: Applications in chemistry
• All-in-one foundational models learning across quantum chemical levels
• Calculated state-of-the art results for solvation and ionization energies of thousands of organic molecules relevant to battery design