Skip to content

Unifying Physical Backpropagation

Aug 2026 · 1 citation
Physics Computer Science

TL;DR

This work identifies sufficient conditions under which the adjoint field required for formally exact gradients can be generated on the same hardware that performs the computation, and shows that reciprocity is a special case of more general intertwining conditions.

Abstract

Physical computing systems exploit device dynamics for computation, but their gradient-based optimization is challenging: backpropagation through a digital twin suffers from a model-reality gap. On-device gradient computation could resolve this issue, and a handful of theoretical and experimental studies have proposed ways to achieve it. Yet a unifying theory identifying when a physical system can compute the gradient of its own performance has been missing. Here we develop such a unification based on the adjoint method: we identify sufficient conditions under which the adjoint field required for formally exact gradients can be generated on the same hardware that performs the computation. Linear and nonlinear systems obey fundamentally different conditions: for linear systems, damping or gain is admissible provided reciprocity is preserved. For nonlinear trajectory systems, the sufficient conditions are reciprocity of the linearized system and the existence of a time-reversal mirror. Algorithmically, the nonlinear case requires infinitesimal nudging, whereas linear systems admit a finite-amplitude experiment. We recover (quantum) Equilibrium Propagation, Hamiltonian echo backpropagation, fully forward mode training and in situ gradient methods in integrated-photonic and free-space-optical systems. Finally, we show that reciprocity is a special case of more general intertwining conditions. For linear systems, these permit exact on-device gradients in a class of non-Hermitian, non-reciprocal systems. For nonlinear trajectories, they combine with generalized time-reversal mirrors to cover, e.g., PT-symmetric equations. The framework also includes time-dependent parameters and Onsager-reciprocal dynamics, providing a unified basis for formally exact physical learning.

View source

Similar papers

Preprint Aug 2026

Physics-Informed Stochastic Configuration Machine: A Backpropagation-Free Neural Network with Fast Training for Nonlinear Differential Equations

The Physics-Informed Stochastic Configuration Machine is proposed, a novel backpropagation-free framework for both forward and inverse problems in differential equations that achieves high-fidelity predictive accuracy and robust parameter identification while accelerating the training process by orders of magnitude com...

Yueze Song, Zhong-Zhe Chen, Li-Hui Cen et al. · 0 citations

Beyond PINNs: A Unified Gauss-Newton and Petrov-Galerkin Framework for Neural and Hybrid PDE Solvers

This work introduces a common framework based on the discretization of functional Gauss--Newton problems by finite families of linear measurements and shows that, through an appropriate duality pairing, the linear measurements can be represented by test functions.

Nilo Schwencke, Roland Maier · 0 citations
Open access Oct 2026

Quantum algorithms for general nonlinear dynamics based on the Carleman embedding

Important nonlinear dynamics, such as those found in plasma and fluid systems, are typically hard to simulate on classical computers. Thus, if fault-tolerant quantum computers could efficiently solve such nonlinear problems, it would be a transformative change for many industries. In a recent breakthrough [Liu et al....

David Jennings, K. Korzekwa, Matteo Lostaglio et al. · 0 citations
Preprint Sep 2026

On the numerical limitations of dual Koopman von Neumann embeddings for solving conservative nonlinear ordinary differential equations on quantum computers

The simulation of nonlinear ordinary differential equations on quantum computers is inherently challenging, as quantum gates are linear operators on qubit states. In this paper, we put forth a Koopman-von Neumann (KvN) operator based algorithm for solving nonlinear ordinary differential equations on a quantum computer...

Thibault Fredon, A. K. Ram, Fabrice Debbasch et al. · 0 citations
Open access Oct 2026

Exact Principal Koopman Eigenfunctions of a Pendulum

The Koopman operator framework offers a rigorous approach to representing nonlinear dynamical systems through linear (or bilinear) differential equations. For deterministic, dissipative systems, it guarantees the existence of global coordinate transformations that linearize autonomous systems and bilinearize input‐af...

U. Römer · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.