[Paper Review] Quantum differential equation solvers: limitations and fast-forwarding

Dong An, Jinpeng Liu+2 more|arXiv (Cornell University)|Nov 9, 2022

Neural Networks and Reservoir Computing25 citations

TL;DR

The paper analyzes quantum algorithms for linear ODEs beyond quantum dynamics, proving lower bounds tied to real-part gaps and non-normality, and develops fast-forwarding techniques for special cases with known eigensystems, including evolution PDEs.

ABSTRACT

We study the limitations and fast-forwarding of quantum algorithms for linear ordinary differential equation (ODE) systems with a particular focus on non-quantum dynamics, where the coefficient matrix in the ODE is not anti-Hermitian or the ODE is inhomogeneous. On the one hand, for generic linear ODEs, by proving worst-case lower bounds, we show that quantum algorithms suffer from computational overheads due to two types of ``non-quantumness'': real part gap and non-normality of the coefficient matrix. We then show that homogeneous ODEs in the absence of both types of ``non-quantumness'' are equivalent to quantum dynamics, and reach the conclusion that quantum algorithms for quantum dynamics work best. To obtain these lower bounds, we propose a general framework for proving lower bounds on quantum algorithms that are amplifiers, meaning that they amplify the difference between a pair of input quantum states. On the other hand, we show how to fast-forward quantum algorithms for solving special classes of ODEs which leads to improved efficiency. More specifically, we obtain exponential improvements in both $T$ and the spectral norm of the coefficient matrix for inhomogeneous ODEs with efficiently implementable eigensystems, including various spatially discretized linear evolutionary partial differential equations. We give fast-forwarding algorithms that are conceptually different from existing ones in the sense that they neither require time discretization nor solving high-dimensional linear systems.

Motivation & Objective

• Identify fundamental limitations (lower bounds) for generic quantum ODE solvers beyond quantum dynamics.
• Characterize conditions under which non-quantumness (real-part gaps, non-normality) hinders efficiency.
• Develop fast-forwarding strategies for tailored ODE solvers with known eigensystems.
• Extend fast-forwarding to inhomogeneous ODEs and apply to spatially discretized PDEs.
• Propose a one-shot solver design that avoids time discretization and high-dimensional linear systems.

Proposed method

• Introduce an amplifier-based framework to derive lower bounds by relating ODE solvers to quantum state discrimination.
• Prove lower bounds showing exponential overhead when real-part gaps exist in eigenvalues.
• Prove linear overhead with non-normal coefficient matrices.
• Show shifting equivalence: shifting A by a real constant aligns homogeneous non-quantum dynamics with quantum dynamics.
• Develop fast-forwarding algorithms for negative semi-definite A with known eigenvalues/eigenvectors and for general matrices with classically computable eigenpairs and quantumly implementable eigenstates.
• Generalize to inhomogeneous ODEs and time-dependent inhomogeneous terms using linear combination of quantum states.

Experimental results

Research questions

• RQ1 Do eigenvalue real-part gaps and matrix non-normality impose fundamental lower bounds on generic quantum ODE solvers?
• RQ2 Can we fast-forward quantum ODE solvers for special classes of A (e.g., with known eigensystems) and/or structured inhomogeneous terms?
• RQ3 How do shifting equivalence and normality relate to equivalence with quantum dynamics?
• RQ4 Can fast-forwarding extend to time-dependent or inhomogeneous terms and to PDE discretizations?
• RQ5 What are practical applications to evolutionary PDEs with spatial discretization?

Key findings

• Low bounds: exponential overhead when eigenvalues have different real parts (real-part gap).
• Low bounds: linear overhead when A is non-normal (based on non-normality measure).
• Homogeneous ODEs with normal A and common real parts are equivalent to quantum dynamics via shifting equivalence.
• Fast-forwarding possible for special A: exponential improvements in time T and/or norm ∥A∥ when eigenvalues/eigenvectors are classically computable and quantumly implementable.
• One-shot, non-time-discretized solvers for inhomogeneous ODEs achieve fast-forwarding by linearly combining homogeneous and inhomogeneous parts.
• Applications to evolutionary PDEs (parabolic, hyperbolic, and higher-order) with known eigensystems enable fast-forwarding.