**Are you an EPFL student looking for a semester project?**

Work with us on data science and visualisation projects, and deploy your project as an app on top of GraphSearch.

Publication# Simultaneous Perturbation Stochastic Approximation of the Quantum Fisher Information

Giuseppe Carleo, Julien Sebastian Gacon, Stefan Woerner

*VEREIN FORDERUNG OPEN ACCESS PUBLIZIERENS QUANTENWISSENSCHAF, *2021

Journal paper

Journal paper

Abstract

The Quantum Fisher Information matrix (QFIM) is a central metric in promising algorithms, such as Quantum Natural Gradient Descent and Variational Quantum Imaginary Time Evolution. Computing the full QFIM for a model with d parameters, however, is computation-ally expensive and generally requires O(d(2)) function evaluations. To remedy these increasing costs in high-dimensional parameter spaces, we propose using simultaneous perturbation stochastic approximation techniques to approximate the QFIM at a constant cost. We present the resulting algorithm and successfully apply it to prepare Hamiltonian ground states and train Variational Quantum Boltzmann Machines.

Official source

This page is automatically generated and may contain information that is not correct, complete, up-to-date, or relevant to your search query. The same applies to every other page on this website. Please make sure to verify the information with EPFL's official sources.

Related concepts

Loading

Related publications

Loading

Related concepts (7)

Related publications (1)

Genetic algorithm

In computer science and operations research, a genetic algorithm (GA) is a metaheuristic inspired by the process of natural selection that belongs to the larger class of evolutionary algorithms (EA)

Algorithm

In mathematics and computer science, an algorithm (ˈælɡərɪðəm) is a finite sequence of rigorous instructions, typically used to solve a class of specific problems or to perform a computation. Algo

Apply

In mathematics and computer science, apply is a function that applies a function to arguments. It is central to programming languages derived from lambda calculus, such as LISP and Scheme, and also i

Loading

Understanding the elementary steps involved in a chemical reaction forms the cornerstone of physical chemistry research. One way to deepen this understanding is by studying chemical and physical processes using linear and nonlinear spectroscopic techniques. However, the outcomes of such experiments can be difficult to decipher due to the interweaving of several effects. Therefore, in order to help experimentalists to disentangle such spectra, the role of theorists is to develop efficient tools that are able to accurately describe molecular systems. The starting point of such tools is solving the time-dependent Schrödinger equation. In this thesis, after implementing geometric integrators, which are based on a combination of the split-operator algorithm and Magnus expansion, for the exact nonadiabatic quantum dynamics of a molecule interacting with a time-dependent electromagnetic field, we derive and implement these geometric integrators for the time-dependent perturbation theory, the Condon, rotating-wave, and ultrashort-pulse approximations, as well as every possible combination thereof. As verified in several model systems, these integrators exactly preserve the geometric invariants, and achieve an arbitrary prescribed order of accuracy in the time step and an exponential convergence in the grid spacing. We also explore in more detail the ultrashort-pulse approximation and derive an analytical expression for the combination with the time-dependent perturbation theory; this expression significantly accelerates numerical calculations. We show that in the limit of the zero pulse width, the d-pulse approximation is recovered. We illustrate the performance of the introduced approximations, using a three-dimensional model of pyrazine, in which it is essential to go beyond the d-pulse limit in order to describe the dynamics correctly. The high-order algorithms are also applied to the photodissociation dynamics of iodomethane (CH3I), following its excitation to the A band. We implement a general split-operator with both discrete-variable and finite-basis representations that can treat one non-Cartesian, such as angular coordinate. To test the effect of various degrees of freedom and of the nonadiabatic dynamics, we apply these algorithms to one-, two-, and three-dimensional models of iodomethane, both in the presence and in the absence of nonadiabatic couplings. A full quantum calculation is, however, limited to problems with low dimensionality (approximately ten degrees of freedom). Beyond this, one must seek an affordable balance between computational efficiency and physical accuracy and can employ, for example, semiclassical methods that are based on classical trajectories. A simple semiclassical approximation that can treat larger systems and requires only local knowledge of the potential is the on-the-fly ab initio thawed Gaussian approximation. We implement a generalization of the method that goes beyond the Franck-Condon approximation and treats Herzberg-Teller active molecules. Our method is used to compute absorption spectra of phenyl radical and of benzene, for which the Herzberg-Teller contribution is essential.