Deep learning density functionals for gradient descent optimization

Abstract

Machine-learned regression models represent a promising tool to implement accurate and computationally affordable energy-density functionals to solve quantum many-body problems via density functional theory. However, while they can easily be trained to accurately map ground-state density profiles to the corresponding energies, their functional derivatives often turn out to be too noisy, leading to instabilities in self-consistent iterations and in gradient-based searches of the ground-state density profile. We investigate how these instabilities occur when standard deep neural networks are adopted as regression models, and we show how to avoid it using an ad-hoc convolutional architecture featuring an inter-channel averaging layer. The testbed we consider is a realistic model for noninteracting atoms in optical speckle disorder. With the inter-channel average, accurate and systematically improvable ground-state energies and density profiles are obtained via gradient-descent optimization, without instabilities nor violations of the variational principle.

arXiv comment9 pages, 10 figures

References in corpus (22)

• Machine learning and the physical sciences
• Anderson localization of a non-interacting Bose-Einstein condensate
• Perspective on density functional theory
• Direct observation of Anderson localization of matter-waves in a controlled disorder
• By-passing the Kohn-Sham equations with machine learning
• Finding Density Functionals with Machine Learning
• Kohn-Sham equations as regularizer: building prior knowledge into machine-learned physics
• Machine Learning for Quantum Matter
• Deep Learning and Density Functional Theory
• Machine learning and density functional theory
• Pure density functional for strong correlations and the thermodynamic limit from machine learning
• Deep learning the Hohenberg-Kohn maps of Density Functional Theory
• Collective dynamics and expansion of a Bose-Einstein condensate in a random potential
• Extensive deep neural networks for transferring small scale learning to large scale systems
• Machine learning density functional theory for the Hubbard model
• Toward Orbital-Free Density Functional Theory with Small Data Sets and Deep Learning
• Supervised machine learning of ultracold atoms with speckle disorder
• Scalable neural networks for the efficient learning of disordered quantum systems
• Efficient Learning of a One-dimensional Density Functional Theory
• Few-boson localization in a continuum with speckle disorder
• Semiclassical spectral function and density of states in speckle potentials
• Supervised learning of few dirty bosons with variable particle number

Cited by in corpus (5)

• Supervised learning of random quantum circuits via scalable neural networks
• Deep learning nonlocal and scalable energy functionals for quantum Ising models
• Leveraging Normalizing Flows for Orbital-Free Density Functional Theory
• Solving deep-learning density functional theory via variational autoencoders
• Challenges and opportunities in the supervised learning of quantum circuit outputs