arXiv subjects

Clara C. Wanjura

Publications and source records attributed to Clara C. Wanjura.

At least 19 recordsLinked to original sources

arXiv2026

Unifying Physical Backpropagation

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.

cond-mat.dis-nn↗

arXiv2026

Non-Hermitian topology in driven-dissipative systems: correspondence with quantum correlations and a resource for entanglement

Directional amplification, in which signals are amplified selectively depending on their propagation direction, is a key resource for quantum information processing and stands in one-to-one correspondence with non-trivial non-Hermitian topology. So far, this correspondence has concerned the mean fields, and thus classical response. Here we turn to the quantum fluctuations, giving access to correlations and entanglement. For phase-preserving amplifiers, we derive analytic expressions for the normal and anomalous correlations, showing that non-trivial topology produces correlations that grow exponentially with the distance between modes and approach the largest values compatible with the uncertainty relations. The associated correlation length diverges at the topological phase transition. Entanglement nonetheless remains local, set by the competition between normalised anomalous correlations and the asymmetry of the mode occupations. For the bosonic Kitaev chain, a phase-sensitive amplifier, the system instead splits into two halves that are internally fully correlated yet mutually uncorrelated. Our work prepares the ground for exploring the quantum properties of non-Hermitian topological systems with state-of-the-art platforms such as cavity optomechanics and superconducting circuits.

cond-mat.mes-hall↗

arXiv2026

Dependence of Equilibrium Propagation Training Success on Network Architecture

The rapid rise of artificial intelligence has led to an unsustainable growth in energy consumption. This has motivated progress in neuromorphic computing and physics-based training of learning machines as alternatives to digital neural networks. Many theoretical studies focus on simple architectures like all-to-all or densely connected layered networks. However, these may be challenging to realize experimentally, e.g. due to connectivity constraints. In this work, we investigate the performance of the widespread physics-based training method of equilibrium propagation for more realistic architectural choices, specifically, locally connected lattices. We train an XY model and explore the influence of architecture on various benchmark tasks, tracking the evolution of spatially distributed responses and couplings during training. Our results show that sparse networks with only local connections can achieve performance comparable to dense networks. Our findings provide guidelines for further scaling up architectures based on equilibrium propagation in realistic settings.

cs.LG↗

arXiv2025

Artificial discovery of lattice models for wave transport

Wave transport devices, such as amplifiers, frequency converters, and nonreciprocal devices, are essential for modern communication, signal processing, and sensing applications. Of particular interest are traveling wave setups, which offer excellent gain and bandwidth properties. So far, the conceptual design of those devices has relied on human ingenuity. This makes it difficult and time-consuming to explore the full design space under a variety of constraints and target functionalities. In our work, we present a method which automates this challenge. By optimizing the discrete and continuous parameters of periodic coupled-mode lattices, our approach identifies the simplest lattices that achieve the target transport functionality, and we apply it to discover new schemes for directional amplifiers, isolators, and frequency demultiplexers. Leveraging automated symbolic regression tools, we find closed analytical expressions that facilitate the discovery of generalizable construction rules. Moreover, we utilize important conceptual connections between the device transport properties and non-Hermitian topology. The resulting structures can be implemented on a variety of platforms, including microwave, optical, and optomechanical systems. Our approach opens the door to extensions like the artificial discovery of lattice models with desired properties in higher dimensions or with nonlinear interactions.

physics.optics↗

arXiv2025

Unifying framework for non-Hermitian and Hermitian topology in driven-dissipative systems

Recently, a one-to-one correspondence between non-trivial non-Hermitian topology and directional amplification has been demonstrated, theoretically and experimentally, for the case of one complex band. Here, we extend our framework to multiple bands and higher spatial dimension. This proves to be far from trivial. Building on the singular value decomposition, we introduce a new quantity that we dub generalised singular spectrum (GSS). The GSS allows us to define physically meaningful bands related to the system's scattering behaviour and to define invariants for novel notions of point gaps (non-Hermitian topology) and line gaps (Hermitian-like topology), respectively. For both invariants, we prove a bulk-boundary correspondence and show that they give rise to two different kinds of topological edge modes. We illustrate our results with a 1D non-Hermitian Su-Schrieffer-Heeger (SSH) model and a 2D non-Hermitian model that features corner-to-corner amplification. Our work is relevant for many state-of-the-art experimental platforms and it sets the stage for applications such as novel directional amplifiers and non-reciprocal sensors.

cond-mat.mes-hall↗

arXiv2025

Training nonlinear optical neural networks with Scattering Backpropagation

As deep learning applications continue to deploy increasingly large artificial neural networks, the associated high energy demands are creating a need for alternative neuromorphic approaches. Optics and photonics are particularly compelling platforms as they offer high speeds and energy efficiency. Neuromorphic systems based on nonlinear optics promise high expressivity with a minimal number of parameters. However, so far, there is no efficient and generic physics-based training method allowing us to extract gradients for the most general class of nonlinear optical systems. In this work, we present Scattering Backpropagation, an efficient method for experimentally measuring approximated gradients for nonlinear optical neural networks. Remarkably, our approach does not require a mathematical model of the physical nonlinearity, and only involves two scattering experiments to extract all gradient approximations. The estimation precision depends on the deviation from reciprocity. We successfully apply our method to well-known benchmarks such as XOR and MNIST. Scattering Backpropagation is widely applicable to existing state-of-the-art, scalable platforms, such as optics, microwave, and also extends to other physical platforms such as electrical circuits.

physics.optics↗

arXiv2024

Detailed balance in non-equilibrium dynamics of granular matter: derivation and implications

Modelling the dynamics of dense granular media is a long standing challenge and essential to many natural phenomena and technological applications. Here, we trace back puzzling experimental observation of detailed-balanced steady states to self-organisation of the neighbour probability distribution. The emergence of detailed balance in non-equilibrium granular dynamics could constitute a major step toward better models of granular media, as well as provide more insight into non-equilibrium processes in general. We show analytically that DBSS emerges when a certain neighbour probability is uniform across the system. This condition leads to a conditional cell order distribution being independent of the condition. We then carry out rotational shear experiments, in which this condition is satisfied, and show that they give rise to robust detailed-balanced steady states. We also show that, when the unconditional cell order distribution maximises the entropy, it is determined by a single constant parameter that is characteristic of all cell transitions. These results illustrate the predictive power of recently proposed evolution equations, which pave the way to simpler models of the dynamics of planar granular systems.

cond-mat.soft↗

arXiv2024

Massive quantum systems as interfaces of quantum mechanics and gravity

The traditional view from particle physics is that quantum gravity effects should only become detectable at extremely high energies and small length scales. Due to the significant technological challenges involved, there has been limited progress in identifying experimentally detectable effects that can be accessed in the foreseeable future. However, in recent decades, the size and mass of quantum systems that can be controlled in the laboratory have reached unprecedented scales, enabled by advances in ground-state cooling and quantum-control techniques. Preparations of massive systems in quantum states pave the way for the explorations of a low-energy regime in which gravity can be both sourced and probed by quantum systems. Such approaches constitute an increasingly viable alternative to accelerator-based, laser-interferometric, torsion-balance, and cosmological tests of gravity. In this review, we provide an overview of proposals where massive quantum systems act as interfaces between quantum mechanics and gravity. We discuss conceptual difficulties in the theoretical description of quantum systems in the presence of gravity, review tools for modeling massive quantum systems in the laboratory, and provide an overview of the current state-of-the-art experimental landscape. Proposals covered in this review include, among others, precision tests of gravity, tests of gravitationally-induced wavefunction collapse and decoherence, as well as gravitymediated entanglement. We conclude the review with an outlook and summary of the key questions raised.

quant-ph↗

arXiv2024

Quantum Equilibrium Propagation for efficient training of quantum systems based on Onsager reciprocity

The widespread adoption of machine learning and artificial intelligence in all branches of science and technology has created a need for energy-efficient, alternative hardware platforms. While such neuromorphic approaches have been proposed and realised for a wide range of platforms, physically extracting the gradients required for training remains challenging as generic approaches only exist in certain cases. Equilibrium propagation (EP) is such a procedure that has been introduced and applied to classical energy-based models which relax to an equilibrium. Here, we show a direct connection between EP and Onsager reciprocity and exploit this to derive a quantum version of EP. This can be used to optimize loss functions that depend on the expectation values of observables of an arbitrary quantum system. Specifically, we illustrate this new concept with supervised and unsupervised learning examples in which the input or the solvable task is of quantum mechanical nature, e.g., the recognition of quantum many-body ground states, quantum phase exploration, sensing and phase boundary exploration. We propose that in the future quantum EP may be used to solve tasks such as quantum phase discovery with a quantum simulator even for Hamiltonians which are numerically hard to simulate or even partially unknown. Our scheme is relevant for a variety of quantum simulation platforms such as ion chains, superconducting qubit arrays, neutral atom Rydberg tweezer arrays and strongly interacting atoms in optical lattices.

quant-ph↗

arXiv2024

Training of Physical Neural Networks

Physical neural networks (PNNs) are a class of neural-like networks that leverage the properties of physical systems to perform computation. While PNNs are so far a niche research area with small-scale laboratory demonstrations, they are arguably one of the most underappreciated important opportunities in modern AI. Could we train AI models 1000x larger than current ones? Could we do this and also have them perform inference locally and privately on edge devices, such as smartphones or sensors? Research over the past few years has shown that the answer to all these questions is likely "yes, with enough research": PNNs could one day radically change what is possible and practical for AI systems. To do this will however require rethinking both how AI models work, and how they are trained - primarily by considering the problems through the constraints of the underlying hardware physics. To train PNNs at large scale, many methods including backpropagation-based and backpropagation-free approaches are now being explored. These methods have various trade-offs, and so far no method has been shown to scale to the same scale and performance as the backpropagation algorithm widely used in deep learning today. However, this is rapidly changing, and a diverse ecosystem of training techniques provides clues for how PNNs may one day be utilized to create both more efficient realizations of current-scale AI models, and to enable unprecedented-scale models.

physics.app-ph↗

arXiv2024

Training Coupled Phase Oscillators as a Neuromorphic Platform using Equilibrium Propagation

Given the rapidly growing scale and resource requirements of machine learning applications, the idea of building more efficient learning machines much closer to the laws of physics is an attractive proposition. One central question for identifying promising candidates for such neuromorphic platforms is whether not only inference but also training can exploit the physical dynamics. In this work, we show that it is possible to successfully train a system of coupled phase oscillators - one of the most widely investigated nonlinear dynamical systems with a multitude of physical implementations, comprising laser arrays, coupled mechanical limit cycles, superfluids, and exciton-polaritons. To this end, we apply the approach of equilibrium propagation, which permits to extract training gradients via a physical realization of backpropagation, based only on local interactions. The complex energy landscape of the XY/ Kuramoto model leads to multistability, and we show how to address this challenge. Our study identifies coupled phase oscillators as a new general-purpose neuromorphic platform and opens the door towards future experimental implementations.

cs.ET↗

arXiv2023

Optomechanical realization of the bosonic Kitaev-Majorana chain

The fermionic Kitaev chain is a canonical model featuring topological Majorana zero modes. We report the experimental realization of its bosonic analogue in a nano-optomechanical network where parametric interactions induce two-mode squeezing and beamsplitter coupling among the nanomechanical modes, equivalent to hopping and superconductor pairing in the fermionic case, respectively. We observe several extraordinary phenomena in the bosonic dynamics and transport, including quadrature-dependent chiral amplification, exponential scaling of the gain with system size, and strong sensitivity to boundary conditions. Controlling the interaction phases and amplitudes uncovers a rich dynamical phase diagram that links the observed phenomena to non-Hermitian topological phase transitions. Finally, we present an experimental demonstration of an exponentially enhanced response to a small perturbation as a consequence of non-Hermitian topology. These results represent the demonstration of a novel synthetic phase of matter whose bosonic dynamics do not have fermionic parallels, and establish a powerful system to study non-Hermitian topology and its applications in signal manipulation and sensing.

quant-ph↗

arXiv2023

Restoration of the non-Hermitian bulk-boundary correspondence via topological amplification

Non-Hermitian (NH) lattice Hamiltonians display a unique kind of energy gap and extreme sensitivity to boundary conditions. Due to the NH skin effect, the separation between edge and bulk states is blurred and the (conventional) bulk-boundary correspondence is lost. Here, we restore the bulk-boundary correspondence for the most paradigmatic class of NH Hamiltonians, namely those with one complex band and without symmetries. We obtain the desired NH Hamiltonian from the (mean-field) unconditional evolution of driven-dissipative cavity arrays, in which NH terms -- in the form of non-reciprocal hopping amplitudes, gain and loss -- are explicitly modeled via coupling to (engineered and non-engineered) reservoirs. This approach removes the arbitrariness in the definition of the topological invariant, as point-gapped spectra differing by a complex-energy shift are not treated as equivalent; the origin of the complex plane provides a common reference (base point) for the evaluation of the topological invariant. This implies that topologically non-trivial Hamiltonians are only a strict subset of those with a point gap and that the NH skin effect does not have a topological origin. We analyze the NH Hamiltonians so obtained via the singular value decomposition, which allows to express the NH bulk-boundary correspondence in the following simple form: an integer value $ν$ of the topological invariant defined in the bulk corresponds to $\vert ν\vert$ singular vectors exponentially localized at the system edge under open boundary conditions, in which the sign of $ν$ determines which edge. Non-trivial topology manifests as directional amplification of a coherent input with gain exponential in system size. Our work solves an outstanding problem in the theory of NH topological phases and opens up new avenues in topological photonics.

quant-ph↗

arXiv2023

Fully Non-Linear Neuromorphic Computing with Linear Wave Scattering

The increasing complexity of neural networks and the energy consumption associated with training and inference create a need for alternative neuromorphic approaches, e.g. using optics. Current proposals and implementations rely on physical non-linearities or opto-electronic conversion to realise the required non-linear activation function. However, there are significant challenges with these approaches related to power levels, control, energy-efficiency, and delays. Here, we present a scheme for a neuromorphic system that relies on linear wave scattering and yet achieves non-linear processing with a high expressivity. The key idea is to inject the input via physical parameters that affect the scattering processes. Moreover, we show that gradients needed for training can be directly measured in scattering experiments. We predict classification accuracies on par with results obtained by standard artificial neural networks. Our proposal can be readily implemented with existing state-of-the-art, scalable platforms, e.g. in optics, microwave and electrical circuits, and we propose an integrated-photonics implementation based on racetrack resonators that achieves high connectivity with a minimal number of waveguide crossings.

physics.optics↗

arXiv2023

Quadrature nonreciprocity: unidirectional bosonic transmission without breaking time-reversal symmetry

Nonreciprocity means that the transmission of a signal depends on its direction of propagation. Despite vastly different platforms and underlying working principles, the realisations of nonreciprocal transport in linear, time-independent systems rely on Aharonov-Bohm interference among several pathways and require breaking time-reversal symmetry. Here we extend the notion of nonreciprocity to unidirectional bosonic transport in systems with a time-reversal symmetric Hamiltonian by exploiting interference between beamsplitter (excitation preserving) and two-mode-squeezing (excitation non-preserving) interactions. In contrast to standard nonreciprocity, this unidirectional transport manifests when the mode quadratures are resolved with respect to an external reference phase. Hence we dub this phenomenon quadrature nonreciprocity. First, we experimentally demonstrate it in the minimal system of two coupled nanomechanical modes orchestrated by optomechanical interactions. Next, we develop a theoretical framework to characterise the class of networks exhibiting quadrature nonreciprocity based on features of their particle-hole graphs. In addition to unidirectionality, these networks can exhibit an even-odd pairing between collective quadratures, which we confirm experimentally in a four-mode system, and an exponential end-to-end gain in the case of arrays of cavities. Our work opens up new avenues for signal routing and quantum-limited amplification in bosonic systems.

cond-mat.mes-hall↗

arXiv2020

Correspondence between non-Hermitian topology and directional amplification in the presence of disorder

In order for non-Hermitian (NH) topological effects to be relevant for practical applications, it is necessary to study disordered systems. In the absence of disorder, certain driven-dissipative cavity arrays with engineered non-local dissipation display directional amplification when associated with a non-trivial winding number of the NH dynamic matrix. In this work, we show analytically that the correspondence between NH topology and directional amplification holds even in the presence of disorder. When a system with non-trivial topology is tuned close to the exceptional point, perfect non-reciprocity (quantified by a vanishing reverse gain) is preserved for arbitrarily strong on-site disorder. For bounded disorder, we derive simple bounds for the probability distribution of the scattering matrix elements. These bounds show that the essential features associated with non-trivial NH topology, namely that the end-to-end forward (reverse) gain grows (is suppressed) exponentially with system size, are preserved in disordered systems. NH topology in cavity arrays is robust and can thus be exploited for practical applications.

cond-mat.mes-hall↗

arXiv2020

Topological framework for directional amplification in driven-dissipative cavity arrays

Directional amplification, in which signals are selectively amplified depending on their propagation direction, has attracted much attention as key resource for applications, including quantum information processing. Recently, several, physically very different, directional amplifiers have been proposed and realized in the lab. In this work, we present a unifying framework based on topology to understand non-reciprocity and directional amplification in driven-dissipative cavity arrays. Specifically, we unveil a one-to-one correspondence between a non-zero topological invariant defined on the spectrum of the dynamic matrix and regimes of directional amplification, in which the end-to-end gain grows exponentially with the number of cavities. We compute analytically the scattering matrix, the gain and reverse gain, showing their explicit dependence on the value of the topological invariant. Parameter regimes achieving directional amplification can be elegantly obtained from a topological `phase diagram', which provides a guiding principle for the design of both phase-preserving and phase-sensitive multimode directional amplifiers.

cond-mat.mes-hall↗

arXiv2019

Structural Evolution of Granular Systems: Theory

A general theory is developed for the evolution of the cell order (CO) distribution in planar granular systems. Dynamic equations are constructed and solved in closed form for several examples: systems under compression; dilation of very dense systems; and the general approach to steady state. We find that all the steady states are stable and that they satisfy detailed balance-like condition when the CO$\,\leq 6$. Illustrative numerical solutions of the evolution are shown. Our theoretical results are validated against an extensive simulation of a sheared system. The formalism can be readily extended to other structural characteristics, paving the way to a general theory of structural organisation of granular systems.

cond-mat.soft↗