arXiv subjects

Masahito Ueda

Publications and source records attributed to Masahito Ueda.

At least 37 records · Page 2Linked to original sources

arXiv2025

Topology of Discrete Quantum Feedback Control

A general framework for analyzing the topology of quantum channels of single-particle systems is developed to find a class of genuinely dynamical topological phases that can be realized by means of discrete quantum feedback control. We provide a symmetry classification of quantum channels by identifying ten symmetry classes of discrete quantum feedback control with projective measurements. We construct various types of topological feedback control by using topological Maxwell demons that achieve robust feedback-controlled chiral or helical transport against noise and decoherence. Topological feedback control thus offers a versatile tool for creating and controlling nonequilibrium topological phases in open quantum systems that are distinct from non-Hermitian and Lindbladian systems and should provide a guiding principle for topology-based design of quantum feedback control.

cond-mat.mes-hall↗

arXiv2025

Universal Upper Bound on Ergotropy and No-Go Theorem by the Eigenstate Thermalization Hypothesis

We show that the maximum extractable work (ergotropy) from a quantum many-body system is constrained by local athermality of an initial state and local entropy decrease brought about by quantum operations. The obtained universal upper bound on ergotropy implies that the eigenstate thermalization hypothesis prohibits work extraction from energy eigenstates by means of finite-time unitary operations. This no-go property implies that Planck's principle, a form of the second law of thermodynamics, holds even for pure quantum states. Our result bridges two independently studied concepts of quantum thermodynamics, the second law and thermalization, via intrasystem correlations in many-body systems as a resource for work extraction.

quant-ph↗

arXiv2024

Exact eigenstates of multicomponent Hubbard models: SU($N$) magnetic $η$ pairing, weak ergodicity breaking, and partial integrability

We construct exact eigenstates of multicomponent Hubbard models in arbitrary dimensions by generalizing the $η$-pairing mechanism. Our models include the SU($N$) Hubbard model as a special case. Unlike the conventional two-component case, the generalized $η$-pairing mechanism permits the construction of eigenstates that feature off-diagonal long-range order and magnetic long-range order. These states form fragmented fermionic condensates due to a simultaneous condensation of multicomponent $η$ pairs. While the $η$-pairing states in the SU(2) Hubbard model are based on the $η$-pairing symmetry, the exact eigenstates in the $N$-component system with $N\geq 3$ arise not from symmetry of the Hamiltonian but from a spectrum generating algebra defined in a Hilbert subspace. We exploit this fact to show that the generalized $η$-pairing eigenstates do not satisfy the eigenstate thermalization hypothesis and serve as quantum many-body scar states. This result indicates a weak breakdown of ergodicity in the $N$-component Hubbard models for $N\geq 3$. Furthermore, we show that these exact eigenstates constitute integrable subsectors in which the Hubbard Hamiltonian effectively reduces to a non-interacting model. This partial integrability causes various multicomponent Hubbard models to weakly break ergodicity. We propose a method of harnessing dissipation to distill the integrable part of the dynamics and elucidate a mechanism of non-thermalization caused by dissipation. This work establishes the coexistence of off-diagonal long-range order and SU($N$) magnetism in excited eigenstates of the multicomponent Hubbard models, which presents a possibility of novel out-of-equilibrium pairing states of multicomponent fermions. These models unveil a unique feature of quantum thermalization of multicomponent fermions, which can experimentally be tested with cold-atom quantum simulators.

cond-mat.str-el↗

arXiv2024

Symbolic Equation Solving via Reinforcement Learning

Machine-learning methods are gradually being adopted in a wide variety of social, economic, and scientific contexts, yet they are notorious for struggling with exact mathematics. A typical example is computer algebra, which includes tasks like simplifying mathematical terms, calculating formal derivatives, or finding exact solutions of algebraic equations. Traditional software packages for these purposes are commonly based on a huge database of rules for how a specific operation (e.g., differentiation) transforms a certain term (e.g., sine function) into another one (e.g., cosine function). These rules have usually needed to be discovered and subsequently programmed by humans. Efforts to automate this process by machine-learning approaches are faced with challenges like the singular nature of solutions to mathematical problems, when approximations are unacceptable, as well as hallucination effects leading to flawed reasoning. We propose a novel deep-learning interface involving a reinforcement-learning agent that operates a symbolic stack calculator to explore mathematical relations. By construction, this system is capable of exact transformations and immune to hallucination. Using the paradigmatic example of solving linear equations in symbolic form, we demonstrate how our reinforcement-learning agent autonomously discovers elementary transformation rules and step-by-step solutions.

cs.LG↗

arXiv2024

Type-II Saddles and Probabilistic Stability of Stochastic Gradient Descent

Characterizing and understanding the dynamics of stochastic gradient descent (SGD) around saddle points remains an open problem. We first show that saddle points in neural networks can be divided into two types, among which the Type-II saddles are especially difficult to escape from because the gradient noise vanishes at the saddle. The dynamics of SGD around these saddles are thus to leading order described by a random matrix product process, and it is thus natural to study the dynamics of SGD around these saddles using the notion of probabilistic stability and the related Lyapunov exponent. Theoretically, we link the study of SGD dynamics to well-known concepts in ergodic theory, which we leverage to show that saddle points can be either attractive or repulsive for SGD, and its dynamics can be classified into four different phases, depending on the signal-to-noise ratio in the gradient close to the saddle.

cs.LG↗

arXiv2024

Pink-noise dynamics in an evolutionary game on a regular graph

We consider an iterated multiplayer prisoner's dilemma game on a square lattice and regular graphs based on the pairwise-Fermi update rule, and obtain heat-maps of the fraction of cooperators and the correlation of neighboring pairs. In the heat-map, there is a mixed region where cooperators and defectors coexist, and in the mixed region the correlation between neighbors is enhanced. Moreover, we observe pink-noise behavior in the mixed region, where the power spectrum can be fitted by a power-law function of frequency. We also find that the pink-noise behavior can be reproduced in a simple random-walk model. In particular, we propose a modified random-walk model which can reproduce not only the pink-noise behavior but also the deviation from it observed in a low-frequency region.

physics.comp-ph↗

arXiv2023

Quasiparticles of Decoherence Processes in Open Quantum Many-Body Systems: Incoherentons

The relaxation dynamics of an open quantum system is determined by the competition between the coherent Hamiltonian dynamics of a system and the dissipative dynamics due to interactions with environments. It is therefore of fundamental interest to understand the transition from the coherent to incoherent regimes. We find that hitherto unrecognized quasiparticles -- incoherentons -- describe this coherent-to-incoherent transition in eigenmodes of a Liouvillian superoperator that governs the dynamics of an open quantum many-body system. Here, an incoherenton is defined as an interchain bound state in an auxiliary ladder system that represents the density matrix of a system. The Liouvillian eigenmodes are classified into groups with different decay rates that reflect the number of incoherentons involved therein. We also introduce a spectral gap -- quantum coherence gap -- that separates the different groups of eigenmodes. We demonstrate the existence of incoherentons in a lattice boson model subject to dephasing, and show that the quantum coherence gap closes when incoherentons are deconfined, which signals a dynamical transition from incoherent relaxation with exponential decay to coherent oscillatory relaxation. Furthermore, we discuss how the decoherence dynamics of quantum many-body systems can be understood in terms of the generation, localization, and diffusion of incoherentons.

quant-ph↗

arXiv2023

Experimental observation of the Yang-Lee quantum criticality in open systems

The Yang-Lee edge singularity was originally studied from the standpoint of mathematical foundations of phase transitions, and its physical demonstration has been of active interest both theoretically and experimentally. However, the presence of an imaginary magnetic field in the Yang-Lee edge singularity has made it challenging to develop a direct observation of the anomalous scaling with negative scaling dimension associated with this critical phenomenon. We experimentally implement an imaginary magnetic field and demonstrate the Yang-Lee edge singularity through a nonunitary evolution governed by a non-Hermitian Hamiltonian in an open quantum system, where a classical system is mapped to a quantum system via the equivalent canonical partition function. In particular, we directly observe the partition function in our experiment using heralded single photons. The nonunitary quantum criticality is identified with the singularity at an exceptional point. We also demonstrate unconventional scaling laws for the finite-temperature dynamics unique to quantum systems.

cond-mat.stat-mech↗

arXiv2023

Yang-Lee Zeros, Semicircle Theorem, and Nonunitary Criticality in Bardeen-Cooper-Schrieffer Superconductivity

Yang and Lee investigated phase transitions in terms of zeros of partition functions, namely, Yang-Lee zeros [Phys. Rev. 87, 404 (1952); Phys. Rev. 87, 410 (1952)]. We show that the essential singularity in the superconducting gap is directly related to the number of roots of the partition function of a BCS superconductor. Those zeros are found to be distributed on a semicircle in the complex plane of the interaction strength due to the Fermi-surface instability. A renormalization-group analysis shows that the semicircle theorem holds for a generic quantum many-body system with a marginal coupling, in sharp contrast with the Lee-Yang circle theorem for the Ising spin system. This indicates that the geometry of Yang-Lee zeros is directly connected to the Fermi-surface instability. Furthermore, we unveil the nonunitary criticality in BCS superconductivity that emerges at each individual Yang-Lee zero due to exceptional points and presents a universality class distinct from that of the conventional Yang-Lee edge singularity.

cond-mat.supr-con↗

arXiv2023

SGD with a Constant Large Learning Rate Can Converge to Local Maxima

Previous works on stochastic gradient descent (SGD) often focus on its success. In this work, we construct worst-case optimization problems illustrating that, when not in the regimes that the previous works often assume, SGD can exhibit many strange and potentially undesirable behaviors. Specifically, we construct landscapes and data distributions such that (1) SGD converges to local maxima, (2) SGD escapes saddle points arbitrarily slowly, (3) SGD prefers sharp minima over flat ones, and (4) AMSGrad converges to local maxima. We also realize results in a minimal neural network-like example. Our results highlight the importance of simultaneously analyzing the minibatch sampling, discrete-time updates rules, and realistic landscapes to understand the role of SGD in deep learning.

cs.LG↗

arXiv2023

What shapes the loss landscape of self-supervised learning?

Prevention of complete and dimensional collapse of representations has recently become a design principle for self-supervised learning (SSL). However, questions remain in our theoretical understanding: When do those collapses occur? What are the mechanisms and causes? We answer these questions by deriving and thoroughly analyzing an analytically tractable theory of SSL loss landscapes. In this theory, we identify the causes of the dimensional collapse and study the effect of normalization and bias. Finally, we leverage the interpretability afforded by the analytical theory to understand how dimensional collapse can be beneficial and what affects the robustness of SSL against data imbalance.

cs.LG↗

arXiv2023

Hermitian and Non-Hermitian Topology from Photon-Mediated Interactions

Light can mediate effective dipole-dipole interactions between atoms or quantum emitters coupled to a common environment. Exploiting them to tailor a desired effective Hamiltonian can have major applications and advance the search for many-body phases. Quantum technologies are mature enough to engineer large photonic lattices with sophisticated structures coupled to quantum emitters. In this context, a fundamental problem is to find general criteria to tailor a photonic environment that mediates a desired effective Hamiltonian of the atoms. Among these criteria, topological properties are of utmost importance since an effective atomic Hamiltonian endowed with a non-trivial topology can be protected against disorder and imperfections. Here, we find general theorems that govern the topological properties (if any) of photon-mediated Hamiltonians in terms of both Hermitian and non-Hermitian topological invariants, thus unveiling a system-bath topological correspondence. The results depend on the number of emitters relative to the number of resonators. For a photonic lattice where each mode is coupled to a single quantum emitter, the Altland-Zirnbauer classification of topological insulators allows us to link the topology of the atoms to that of the photonic bath: we unveil the phenomena of topological preservation and reversal to the effect that the atomic topology can be the same or opposite to the photonic one, depending on Hermiticity of the photonic system and on the parity of the spatial dimension. As a consequence, the bulk-edge correspondence implies the existence of atomic boundary modes with the group velocity opposite to the photonic ones in a 2D Hermitian topological system. If there are fewer emitters than photonic modes, the atomic system is less constrained and no general photon-atom topological correspondence can be found. We show this with two counterexamples.

quant-ph↗

arXiv2022

Nonlinear Landauer formula: Nonlinear response theory of disordered and topological materials

The Landauer formula provides a general scattering formulation of electrical conduction. Despite its utility, it has been mainly applied to the linear-response regime, and a scattering theory of nonlinear response has yet to be fully developed. Here, we extend the Landauer formula to the nonlinear-response regime. We show that while the linear conductance is directly related to the transmission probability, the nonlinear conductance is given by its derivatives with respect to energy. This sensitivity to the energy derivatives is shown to produce unique nonlinear transport phenomena of mesoscopic systems including disordered and topological materials. By way of illustration, we investigate nonlinear conductance of disordered chains and identify their universal behavior according to symmetry. In particular, we find large singular nonlinear conductance for zero modes, including Majorana zero modes in topological superconductors. We also show the critical behavior of nonlinear response around the mobility edges due to the Anderson transitions. Moreover, we study nonlinear response of graphene as a prime example of topological materials featuring quantum anomaly. Furthermore, considering the geometry of electronic wave functions, we develop a scattering theory of the nonlinear Hall effect. We establish a new connection between the nonlinear Hall response and the nonequilibrium quantum fluctuations. We also discuss the influence of disorder and Anderson localization on the nonlinear Hall effect. Our work opens a new avenue in quantum physics beyond the linear-response regime.

cond-mat.mes-hall↗

arXiv2022

Three Learning Stages and Accuracy-Efficiency Tradeoff of Restricted Boltzmann Machines

Restricted Boltzmann Machines (RBMs) offer a versatile architecture for unsupervised machine learning that can in principle approximate any target probability distribution with arbitrary accuracy. However, the RBM model is usually not directly accessible due to its computational complexity, and Markov-chain sampling is invoked to analyze the learned probability distribution. For training and eventual applications, it is thus desirable to have a sampler that is both accurate and efficient. We highlight that these two goals generally compete with each other and cannot be achieved simultaneously. More specifically, we identify and quantitatively characterize three regimes of RBM learning: independent learning, where the accuracy improves without losing efficiency; correlation learning, where higher accuracy entails lower efficiency; and degradation, where both accuracy and efficiency no longer improve or even deteriorate. These findings are based on numerical experiments and heuristic arguments.

cs.LG↗

arXiv2022

Lindbladian Many-Body Localization

We discover a novel localization transition that alters the dynamics of coherence in disordered many-body spin systems subject to Markovian dissipation. The transition occurs in the middle spectrum of the Lindbladian super-operator whose eigenstates obey the universality of non-Hermitian random-matrix theory for weak disorder and exhibit localization of off-diagonal degrees of freedom for strong disorder. This Lindbladian many-body localization prevents many-body decoherence due to interactions and is conducive to robustness of the coherent dynamics characterized by the rigidity of the decay rate of coherence.

cond-mat.dis-nn↗

arXiv2022

Universal Thermodynamic Uncertainty Relation in Non-Equilibrium Dynamics

We derive a universal thermodynamic uncertainty relation (TUR) that applies to an arbitrary observable in a general Markovian system. The generality of our result allows us to make two findings: (1) for an arbitrary out-of-equilibrium system, both the entropy production and the \textit{degree of non-stationarity} are required to tightly bound the strength of a thermodynamic current; (2) by removing the antisymmetric constraint on observables, the TUR in physics and a fundamental inequality in theoretical finance can be unified in a single framework.

cond-mat.stat-mech↗

arXiv2022

Embedding the Yang-Lee Quantum Criticality in Open Quantum Systems

The Yang-Lee edge singularity is a quintessential nonunitary critical phenomenon accompanied by anomalous scaling laws. However, an imaginary magnetic field involved in this critical phenomenon makes its physical implementation difficult. By invoking the quantum-classical correspondence to embed the Yang-Lee edge singularity in a quantum system with an ancilla qubit, we demonstrate a physical realization of the nonunitary quantum criticality in an open quantum system. Here the nonunitary criticality is identified with the singularity at an exceptional point caused by postselection of quantum measurement.

cond-mat.stat-mech↗

arXiv2022

Exact Phase Transitions in Deep Learning

This work reports deep-learning-unique first-order and second-order phase transitions, whose phenomenology closely follows that in statistical physics. In particular, we prove that the competition between prediction error and model complexity in the training loss leads to the second-order phase transition for nets with one hidden layer and the first-order phase transition for nets with more than one hidden layer. The proposed theory is directly relevant to the optimization of neural networks and points to an origin of the posterior collapse problem in Bayesian deep learning.

cs.LG↗