arXiv subjects

Xu Zhang

Publications and source records attributed to Xu Zhang.

At least 361 records · Page 20Linked to original sources

arXiv2022

Thermodynamic characteristic for correlated flat-band system with quantum anomalous Hall ground state

While the ground state phase diagram of the correlated flat-band systems have been intensively investigated, the dynamic and thermodynamic properties of such lattice models are less explored, but it is the latter which is most relevant to the experimental probes (transport, quantum capacitance and spectroscopy) of the quantum moiré materials such as twisted bilayer graphene and transition metal dichalcogenides. Here we show, by means of momentum-space quantum Monte Carlo and exact diagonalization, there exists a unique thermodynamic characteristic for the correlated flat-band models with interaction-driven quantum anomalous Hall (QAH) ground state, namely, the transition from the QAH insulator to the metallic state takes place at a much lower temperature compared with the zero-temperature single-particle gap generated by the long-range Coulomb interaction. Such low transition temperature comes from the proliferation of excitonic particle-hole excitations, which "quantum teleport" the electrons across the gap between different topological bands to restore the broken time-reversal symmetry and give rise to a pronounced enhancement in the charge compressibility. Future experiments, to verify such generic thermodynamic characteristics, are proposed.

cond-mat.str-el↗

arXiv2022

Selective Fairness in Recommendation via Prompts

Recommendation fairness has attracted great attention recently. In real-world systems, users usually have multiple sensitive attributes (e.g. age, gender, and occupation), and users may not want their recommendation results influenced by those attributes. Moreover, which of and when these user attributes should be considered in fairness-aware modeling should depend on users' specific demands. In this work, we define the selective fairness task, where users can flexibly choose which sensitive attributes should the recommendation model be bias-free. We propose a novel parameter-efficient prompt-based fairness-aware recommendation (PFRec) framework, which relies on attribute-specific prompt-based bias eliminators with adversarial training, enabling selective fairness with different attribute combinations on sequential recommendation. Both task-specific and user-specific prompts are considered. We conduct extensive evaluations to verify PFRec's superiority in selective fairness. The source codes are released in \url{https://github.com/wyqing20/PFRec}.

cs.IR↗

arXiv2022

Fermion sign bounds theory in quantum Monte Carlo simulation

Sign problem in fermion quantum Monte Carlo (QMC) simulation appears to be an extremely hard problem. Traditional lore passing around for years tells people that when there is a sign problem, the average sign in QMC simulation approaches zero exponentially fast with the space-time volume of the configurational space. We, however, analytically show this is not always the case and manage to find physical bounds for the average sign. Our understanding is based on a direct connection between the sign bounds and a well-defined partition function of reference system and could distinguish when the bounds have the usual exponential scaling, and when they are bestowed on an algebraic scaling at low temperature limit. We analytically explain such algebraic sign problems found in flat band moiré lattice models at low temperature limit. At finite temperature, a domain size argument based on sign bounds also explains the connection between sign behavior and finite temperature phase transition. Sign bounds, as a well-defined observable, may have ability to ease or even make use of the sign problem.

cond-mat.str-el↗

arXiv2022

Data-Time Tradeoffs for Optimal k-Thresholding Algorithms in Compressed Sensing

Optimal $k$-thresholding algorithms are a class of $k$-sparse signal recovery algorithms that overcome the shortcomings of traditional hard thresholding algorithms caused by the oscillation of the residual function. In this paper, a novel convergence analysis for optimal $k$-thresholding algorithms is established, which reveals the data-time tradeoffs of these algorithms. Both the analysis and numerical results demonstrate that when the number of measurements is small, the algorithms cannot converge; when the number of measurements is suitably large, the number of iterations required for successful recovery has a negative correlation with the number of measurements, and the algorithms can achieve linear convergence. Furthermore, the main theorems indicate that the number of measurements required for successful recovery is of the order of $k \log({n}/{k})$, where $n$ is the dimension of the target signal.

cs.IT↗

arXiv2022

Personalized Federated Learning via Variational Bayesian Inference

Federated learning faces huge challenges from model overfitting due to the lack of data and statistical diversity among clients. To address these challenges, this paper proposes a novel personalized federated learning method via Bayesian variational inference named pFedBayes. To alleviate the overfitting, weight uncertainty is introduced to neural networks for clients and the server. To achieve personalization, each client updates its local distribution parameters by balancing its construction error over private data and its KL divergence with global distribution from the server. Theoretical analysis gives an upper bound of averaged generalization error and illustrates that the convergence rate of the generalization error is minimax optimal up to a logarithmic factor. Experiments show that the proposed method outperforms other advanced personalized methods on personalized models, e.g., pFedBayes respectively outperforms other SOTA algorithms by 1.25%, 0.42% and 11.71% on MNIST, FMNIST and CIFAR-10 under non-i.i.d. limited data.

cs.LG↗

arXiv2022

Submanifolds of some Hartogs domain and the complex Euclidean space

Two Kahler manifolds are called relatives if they admit a common Kahler submanifold with the same induced metrics. In this paper, we show that a Hartogs domain over an irreducible bounded symmetric domain equipped with the Bergman metric is not a relative to the complex Euclidean space. This generalizes the results in [5, 4] and the novelty here is that the Bergman kernel of the Hartogs domain is not necessarily Nash algebraic.

math.CV↗

arXiv2022

Topological and Algebraic Structures of Atanassov's Intuitionistic Fuzzy-Values Space

We prove that the space of intuitionistic fuzzy values (IFVs) with a linear order based on a score function and an accuracy function has the same algebraic structure as the one induced by a linear order based on a similarity function and an accuracy function. By introducing a new operator for IFVs via the linear order based on a score function and an accuracy function, we show that such an operator is a strong negation on IFVs. Moreover, we observe that the space of IFVs is a complete lattice and a Kleene algebra with the new operator. We also demonstrate that the topological space of IFVs with the order topology induced by the above two linear orders is not separable and metrizable but compact and connected. From some new perspectives,our results partially answer three open problems posed by Atanassov [Intuitionistic Fuzzy Sets: Theory and Applications, Springer, 1999] and [On Intuitionistic Fuzzy Sets Theory, Springer, 2012]. Furthermore, we construct an isomorphism between the spaces of IFVs and q-rung orthopedic fuzzy values (q-ROFVs) under the corresponding linear orders. To this end, we introduce the concept of admissible similarity measures with particular orders for IFSs, extending the existing definition of the similarity measure for IFSs, and construct an admissible similarity measure with a linear order based on a score function and an accuracy function, which is effectively applied to a pattern recognition problem about the classification of building materials.

cs.AI↗

arXiv2022

Robust single-sideband-modulated Raman light generation for atom interferometry by FBG-based optical rectangular filtration

Low-phase-noise and pure-spectrum Raman light is vital for high-precision atom interferometry by two-photon Raman transition. A preferred and prevalent solution for Raman light generation is electro-optic phase modulation. However, phase modulation inherently brings in double sidebands, resulting in residual sideband effects of multiple laser pairs beside Raman light in atom interferometry. Based on a well-designed rectangular fiber Bragg grating and an electro-optic modulator, optical single-sideband modulation has been realized at 1560 nm with a stable suppression ratio better than -25 dB despite of intense temperature variations. After optical filtration and frequency doubling, a robust phase-coherent Raman light at 780 nm is generated with a stable SNR of better than -19 dB and facilitates measuring the local gravity successfully. This proposed all-fiber single-sideband-modulated Raman light source, characterized as robust, compact and low-priced, is practical and potential for field applications of portable atom interferometry.

physics.atom-ph↗

arXiv2022

Experimental quantum adversarial learning with programmable superconducting qubits

Quantum computing promises to enhance machine learning and artificial intelligence. Different quantum algorithms have been proposed to improve a wide spectrum of machine learning tasks. Yet, recent theoretical works show that, similar to traditional classifiers based on deep classical neural networks, quantum classifiers would suffer from the vulnerability problem: adding tiny carefully-crafted perturbations to the legitimate original data samples would facilitate incorrect predictions at a notably high confidence level. This will pose serious problems for future quantum machine learning applications in safety and security-critical scenarios. Here, we report the first experimental demonstration of quantum adversarial learning with programmable superconducting qubits. We train quantum classifiers, which are built upon variational quantum circuits consisting of ten transmon qubits featuring average lifetimes of 150 $μ$s, and average fidelities of simultaneous single- and two-qubit gates above 99.94% and 99.4% respectively, with both real-life images (e.g., medical magnetic resonance imaging scans) and quantum data. We demonstrate that these well-trained classifiers (with testing accuracy up to 99%) can be practically deceived by small adversarial perturbations, whereas an adversarial training process would significantly enhance their robustness to such perturbations. Our results reveal experimentally a crucial vulnerability aspect of quantum learning systems under adversarial scenarios and demonstrate an effective defense strategy against adversarial attacks, which provide a valuable guide for quantum artificial intelligence applications with both near-term and future quantum devices.

quant-ph↗

arXiv2022

Robust remote estimation over the collision channel in the presence of an intelligent jammer

We consider a sensor-receiver pair communicating over a wireless channel in the presence of a jammer who may launch a denial-of-service attack. We formulate a zero-sum game between a coordinator that jointly designs the transmission and estimation policies, and the jammer. We consider two cases depending on whether the jammer can sense the channel or not. We characterize a saddle-point equilibrium for the class of symmetric and unimodal probability density functions when the jammer cannot sense the channel. If the jammer can sense if the channel is being used, we provide an efficient algorithm that alternates between iterations of Projected Gradient Ascent and the Convex-Concave Procedure to find approximate First-order Nash-Equilibria. Our numerical results show that in certain cases the jammer may decide to launch a denial-of-service attack with the goal of deceiving the receiver even when the sensor decides not to transmit.

eess.SY↗

arXiv2022

Multi-view Multi-behavior Contrastive Learning in Recommendation

Multi-behavior recommendation (MBR) aims to jointly consider multiple behaviors to improve the target behavior's performance. We argue that MBR models should: (1) model the coarse-grained commonalities between different behaviors of a user, (2) consider both individual sequence view and global graph view in multi-behavior modeling, and (3) capture the fine-grained differences between multiple behaviors of a user. In this work, we propose a novel Multi-behavior Multi-view Contrastive Learning Recommendation (MMCLR) framework, including three new CL tasks to solve the above challenges, respectively. The multi-behavior CL aims to make different user single-behavior representations of the same user in each view to be similar. The multi-view CL attempts to bridge the gap between a user's sequence-view and graph-view representations. The behavior distinction CL focuses on modeling fine-grained differences of different behaviors. In experiments, we conduct extensive evaluations and ablation tests to verify the effectiveness of MMCLR and various CL tasks on two real-world datasets, achieving SOTA performance over existing baselines. Our code will be available on \url{https://github.com/wyqing20/MMCLR}

cs.IR↗

arXiv2022

Many-body Hilbert space scarring on a superconducting processor

Quantum many-body scarring (QMBS) -- a recently discovered form of weak ergodicity breaking in strongly-interacting quantum systems -- presents opportunities for mitigating thermalization-induced decoherence in quantum information processsing. However, the existing experimental realizations of QMBS are based on kinetically-constrained systems where an emergent dynamical symmetry "shields" such states from the thermalizing bulk of the spectrum. Here, we experimentally realize a distinct kind of QMBS phenomena by approximately decoupling a part of the many-body Hilbert space in the computational basis. Utilizing a programmable superconducting processor with 30 qubits and tunable couplings, we realize Hilbert space scarring in a non-constrained model in different geometries, including a linear chain as well as a quasi-one-dimensional comb geometry. By performing full quantum state tomography on 4-qubit subsystems, we provide strong evidence for QMBS states by measuring qubit population dynamics, quantum fidelity and entanglement entropy following a quench from initial product states. Our experimental findings broaden the realm of QMBS mechanisms and pave the way to exploiting correlations in QMBS states for applications in quantum information technology.

quant-ph↗

arXiv2022

Dynamical properties of collective excitations in twisted bilayer Graphene

Employing the recently developed momentum-space quantum Monte Carlo scheme, we study the dynamic response of single-particle and collective excitations in realistic continuum models of twisted bilayer graphene. At charge neutrality, this unbiased numerical method reveals strong competition between different symmetry breaking channels with a leading instability towards the intervalley coherent state. Single-particle spectra indicate that repulsive interactions push the fermion spectral weight away from the Fermi energy and open up an insulating gap. The spectra of collective excitations suggest an approximate valley $SU(2)$ symmetry. At low-energy, long-lived valley waves are observed, which resemble spin waves of Heisenberg ferromagnetism. At high-energy, these sharp modes quickly become over-damped, when their energy reaches the fermion particle-hole continuum.

cond-mat.str-el↗

arXiv2022

In-situ probing and stabilizing the power ratio of electro-optic-modulated laser pairs based on VIPA etalon for quantum sensing

Monitoring and stabilizing the power ratio of laser pairs is significant to high-precision atom interferometers, especially as the compact electro-optic modulated all-fiber laser system prevails. In this Letter, we demonstrate a novel method to in-situ probe the relative power of laser pairs and to stabilize the power ratio of two Raman lasers using a high-dispersion virtually imaged phased array (VIPA) etalon. Sub-microsecond resolution on probing laser power transformation during atom interferometer sequence is achieved and the power ratio of two Raman lasers (PRTR) is tightly locked with high bandwidth despite of environmental disturbances, showing an Allan deviation of $4.39\times 10^{-5}$ at 1000 s averaging time. This method provides a novel way to stabilize the PRTR and diagnose the multi-frequency laser systems for atom interferometers and could find potential application in broad quantum sensing scenarios.

physics.atom-ph↗

arXiv2022

A generalized family of transcendental functions with one dimensional Julia sets

A generalized family of transcendental (non-polynomial entire) functions is constructed, where the Hausdorff dimension and the packing dimension of the Julia sets are equal to one. Further, there exist multiply connected wandering domains, the dynamics can be completed described, and for any $s\in(0,+\infty]$, there is a function taken from this family with the order of growth $s$. Baker proved that the Hausdorff dimension of the transcendental function is no less than one in 1975, the minimum value was obtained via an elegant construction by Bishop in 2018. The order of growth is zero in Bishop's construction, the family of functions here have arbitrarily positive or even infinite order of growth.

math.DS↗

arXiv2022

UniParser: A Unified Log Parser for Heterogeneous Log Data

Logs provide first-hand information for engineers to diagnose failures in large-scale online service systems. Log parsing, which transforms semi-structured raw log messages into structured data, is a prerequisite of automated log analysis such as log-based anomaly detection and diagnosis. Almost all existing log parsers follow the general idea of extracting the common part as templates and the dynamic part as parameters. However, these log parsing methods, often neglect the semantic meaning of log messages. Furthermore, high diversity among various log sources also poses an obstacle in the generalization of log parsing across different systems. In this paper, we propose UniParser to capture the common logging behaviours from heterogeneous log data. UniParser utilizes a Token Encoder module and a Context Encoder module to learn the patterns from the log token and its neighbouring context. A Context Similarity module is specially designed to model the commonalities of learned patterns. We have performed extensive experiments on 16 public log datasets and our results show that UniParser outperperforms state-of-the-art log parsers by a large margin.

cs.SE↗

arXiv2022

A two-step backward compatible fullband speech enhancement system

Speech enhancement methods based on deep learning have surpassed traditional methods. While many of these new approaches are operating on the wideband (16kHz) sample rate, a new fullband (48kHz) speech enhancement system is proposed in this paper. Compared to the existing fullband systems that utilizes perceptually motivated features to train the fullband speech enhancement using a single network structure, the proposed system is a two-step system ensuring good fullband speech enhancement quality while backward compatible to the existing wideband systems.

eess.AS↗

arXiv2022

SMDT: Selective Memory-Augmented Neural Document Translation

Existing document-level neural machine translation (NMT) models have sufficiently explored different context settings to provide guidance for target generation. However, little attention is paid to inaugurate more diverse context for abundant context information. In this paper, we propose a Selective Memory-augmented Neural Document Translation model to deal with documents containing large hypothesis space of the context. Specifically, we retrieve similar bilingual sentence pairs from the training corpus to augment global context and then extend the two-stream attention model with selective mechanism to capture local context and diverse global contexts. This unified approach allows our model to be trained elegantly on three publicly document-level machine translation datasets and significantly outperforms previous document-level NMT models.

cs.CL↗