Neural network approximation of regularized density functionals

• URL: http://arxiv.org/abs/2511.18512v1
• Date: Sun, 23 Nov 2025 16:08:24 GMT
• Title: Neural network approximation of regularized density functionals
• Authors: Mihály A. Csirik, Andre Laestadius, Mathias Oster,
• Abstract summary: Density functional theory is one of the most efficient and widely used computational methods of quantum mechanics.<n>We propose a procedure by first applying Moreau-Yosida regularization to make the exact functionals continuous.<n>The resulting neural network preserves the positivity and convexity of the exact functionals.
• Score: 1.2744523252873352
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Density functional theory is one of the most efficient and widely used computational methods of quantum mechanics, especially in fields such as solid state physics and quantum chemistry. From the theoretical perspecive, its central object is the universal density functional which contains all intrinsic information about the quantum system in question. Once the external potential is provided, in principle one can obtain the exact ground-state energy via a simple minimization. However, the universal density functional is a very complicated mathematical object and almost always it is replaced with its approximate variants. So far, no ``first principles'', mathematically consistent and convergent approximation procedure has been devised that has general applicability. In this paper, we propose such a procedure by first applying Moreau--Yosida regularization to make the exact functionals continuous (even differentiable) and then approximate the regularized functional by a neural network. The resulting neural network preserves the positivity and convexity of the exact functionals. More importantly, it is differentiable, so it can be directly used in a Kohn--Sham calculation.

Related papers

• Coupled Cluster con MōLe: Molecular Orbital Learning for Neural Wavefunctions [34.53033480184221]
  Density functional theory (DFT) is the most widely used method for calculating molecular properties.<n> Coupled-cluster (CC) theory is the most successful method for achieving accuracy beyond DFT.<n>We present the Molecular Orbital Learning (MLe) architecture, an equivariant machine learning model that directly predicts CC's core mathematical objects.
  arXiv  Detail & Related papers  (2026-02-23T18:54:46Z)
• A Joint Quantum Computing, Neural Network and Embedding Theory Approach for the Derivation of the Universal Functional [0.0]
  We introduce a novel approach that exploits the intersection of quantum computing, machine learning and reduced density matrix functional theory.<n>Our method focuses on obtaining the universal functional using a deep neural network trained with quantum algorithms.
  arXiv  Detail & Related papers  (2025-12-15T09:46:54Z)
• Geometry of Generalized Density Functional Theories [11.36271800513876]
  Density functional theory (DFT) is an indispensable ab initio method in both quantum chemistry and condensed matter physics.<n>We construct a mathematical framework generalizing all ground state functional theories.<n>We show that within the special class of such functional theories where the space of external potentials forms the Lie algebra of a compact Lie group, the $N$-representability problem is readily solved.
  arXiv  Detail & Related papers  (2025-11-18T15:40:25Z)
• The Integral Decimation Method for Quantum Dynamics and Statistical Mechanics [44.99833362998488]
  A direct numerical evaluation of an multidimensional integral incurs a computational cost that is exponential in the number of dimensions.<n>Here, we derive and implement a quantum algorithm to compress multidimensional integrand into a product of matrix-valued functions.
  arXiv  Detail & Related papers  (2025-06-12T22:26:12Z)
• Highly Accurate Real-space Electron Densities with Neural Networks [7.176850154835262]
  We introduce a novel method to obtain accurate densities from real-space many-electron wave functions. We use variational quantum Monte Carlo with deep-learning ans"atze (deep QMC) to obtain highly accurate wave functions free of basis set errors.
  arXiv  Detail & Related papers  (2024-09-02T14:56:22Z)
• Physics-Informed Neural Networks for an optimal counterdiabatic quantum computation [32.73124984242397]
  We introduce a novel methodology that leverages the strength of Physics-Informed Neural Networks (PINNs) to address the counterdiabatic (CD) protocol in the optimization of quantum circuits comprised of systems with $N_Q$ qubits. The main applications of this methodology have been the $mathrmH_2$ and $mathrmLiH$ molecules, represented by a 2-qubit and 4-qubit systems employing the STO-3G basis.
  arXiv  Detail & Related papers  (2023-09-08T16:55:39Z)
• Orbital-Free Quasi-Density Functional Theory [0.0]
  Wigner functions are broadly used to probe non-classical effects in the macroscopic world. We develop an orbital-free functional framework to compute the 1-body Wigner quasi-probability for both fermionic and bosonic systems.
  arXiv  Detail & Related papers  (2023-04-18T15:27:52Z)
• D4FT: A Deep Learning Approach to Kohn-Sham Density Functional Theory [79.50644650795012]
  We propose a deep learning approach to solve Kohn-Sham Density Functional Theory (KS-DFT) We prove that such an approach has the same expressivity as the SCF method, yet reduces the computational complexity. In addition, we show that our approach enables us to explore more complex neural-based wave functions.
  arXiv  Detail & Related papers  (2023-03-01T10:38:10Z)
• Quantum Geometry of Expectation Values [1.261852738790008]
  We show that the boundary of expectation value space corresponds to the ground state, which presents a natural bound that generalizes Heisenberg's uncertainty principle. Our approach provides an alternative time-independent quantum formulation that transforms the linear problem in a high-dimensional Hilbert space into a nonlinear algebro-geometric problem in a low dimension.
  arXiv  Detail & Related papers  (2023-01-14T14:01:41Z)
• Symbolic Recovery of Differential Equations: The Identifiability Problem [52.158782751264205]
  Symbolic recovery of differential equations is the ambitious attempt at automating the derivation of governing equations. We provide both necessary and sufficient conditions for a function to uniquely determine the corresponding differential equation. We then use our results to devise numerical algorithms aiming to determine whether a function solves a differential equation uniquely.
  arXiv  Detail & Related papers  (2022-10-15T17:32:49Z)
• Density functionals and Kohn-Sham potentials with minimal wavefunction preparations on a quantum computer [0.0]
  One of the potential applications of a quantum computer is solving quantum chemical systems. We demonstrate a method for obtaining the exact functional as a machine learned model from a sufficiently powerful quantum computer.
  arXiv  Detail & Related papers  (2020-08-12T22:50:39Z)
• Variational Monte Carlo calculations of $\mathbf{A\leq 4}$ nuclei with an artificial neural-network correlator ansatz [62.997667081978825]
  We introduce a neural-network quantum state ansatz to model the ground-state wave function of light nuclei. We compute the binding energies and point-nucleon densities of $Aleq 4$ nuclei as emerging from a leading-order pionless effective field theory Hamiltonian.
  arXiv  Detail & Related papers  (2020-07-28T14:52:28Z)
• UNIPoint: Universally Approximating Point Processes Intensities [125.08205865536577]
  We provide a proof that a class of learnable functions can universally approximate any valid intensity function. We implement UNIPoint, a novel neural point process model, using recurrent neural networks to parameterise sums of basis function upon each event.
  arXiv  Detail & Related papers  (2020-07-28T09:31:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.