Automatic differentiation of dominant eigensolver and its applications in quantum physics

Abstract

We investigate the automatic differentiation of dominant eigensolver where only a small proportion of eigenvalues and corresponding eigenvectors are obtained. Backpropagation through the dominant eigensolver involves solving certain low-rank linear systems without direct access to the full spectrum of the problem. Furthermore, the backward pass can be conveniently differentiated again, which implies that in principle one can obtain arbitrarily higher order derivatives of the dominant eigen-decomposition process. These results allow for the construction of an efficient dominant eigensolver primitive, which has wide applications in quantum physics. As a demonstration, we compute second order derivative of the ground state energy and fidelity susceptibility of 1D transverse field Ising model through the exact diagonalization approach. We also calculate the ground state energy of the same model in the thermodynamic limit by performing gradient-based optimization of uniform matrix product states. By programming these computation tasks in a fully differentiable way, one can efficiently handle the dominant eigen-decomposition of very large matrices while still sharing various advantages of differentiable programming paradigm, notably the generic nature of the implementation and free of tedious human efforts of deriving gradients analytically.

15 pages, 5 figures, revised version, code: https://github.com/buwantaiji/DominantSparseEigenAD

References in corpus (29)

• Anyons in an exactly solved model and beyond
• The density-matrix renormalization group in the age of matrix product states
• Paired states of fermions in two dimensions with breaking of parity and time-reversal symmetries, and the fractional quantum Hall effect
• Area laws for the entanglement entropy - a review
• A Practical Introduction to Tensor Networks: Matrix Product States and Projected Entangled Pair States
• Many-body localization edge in the random-field Heisenberg chain
• Classical simulation of infinite-size quantum lattice systems in one spatial dimension
• Ground state overlap and quantum phase transitions
• Matrix product states represent ground states faithfully
• Classical simulation of infinite-size quantum lattice systems in two spatial dimensions
• Tensor renormalization group approach to 2D classical lattice models
• Fidelity approach to quantum phase transitions
• Quantum critical scaling of the geometric tensors
• Corner Transfer Matrix Renormalization Group Method
• Gradient-based Hyperparameter Optimization through Reversible Learning
• Variational optimization algorithms for uniform matrix product states
• Differentiable Programming Tensor Networks
• Quantum Critical Scaling of Fidelity Susceptibility
• Tangent-space methods for uniform matrix product states
• Entropy and Entanglement in Quantum Ground States
• Quantum Monte Carlo simulations of fidelity at magnetic quantum phase transitions
• Speedup for quantum optimal control from automatic differentiation based on graphics processing units
• Fidelity analysis of topological quantum phase transitions
• Automatic differentiation in quantum chemistry with an application to fully variational Hartree-Fock
• Fidelity susceptibility made simple: A unified quantum Monte Carlo approach
• Scaling dimension of fidelity susceptibility in quantum phase transitions
• Fidelity in topological quantum phases of matter
• Construction of Hamiltonians by supervised learning of energy and entanglement spectra
• Backpropagation-Friendly Eigendecomposition

Cited by in corpus (19)

• Kohn-Sham equations as regularizer: building prior knowledge into machine-learned physics
• Investigation of the Néel phase of the frustrated Heisenberg antiferromagnet by differentiable symmetric tensor networks
• Heavy quark potential in quark-gluon Plasma: Deep neural network meets lattice quantum chromodynamics
• A beginner's guide to non-abelian iPEPS for correlated fermions
• Protocol Discovery for the Quantum Control of Majoranas by Differentiable Programming and Natural Evolution Strategies
• Variational optimization of tensor-network states with the honeycomb-lattice corner transfer matrix
• An introduction to infinite projected entangled-pair state methods for variational ground state simulations using automatic differentiation
• Mitigating the fermion sign problem by automatic differentiation
• Automatic Differentiable Numerical Renormalization Group
• Inverse Hamiltonian design by automatic differentiation
• Differentiable programming tensor networks for Kitaev magnets
• Quantum computing fidelity susceptibility using automatic differentiation
• Variational Tensor Network Operator
• Two Dimensional Isometric Tensor Networks on an Infinite Strip
• Unsupervised Learning of Effective Quantum Impurity Models
• Single-layer framework of variational tensor network states
• Differentiate Everything with a Reversible Embeded Domain-Specific Language
• Preparation of Many-body Ground States by Time Evolution with Variational Microscopic Magnetic Fields and Incomplete Interactions
• Learning holographic QCD with unflavored meson spectra