Searcharxiv⌕ Search

arXiv subjects

Guangyu Wu

Publications and source records attributed to Guangyu Wu.

18 recordsLinked to original sources
arXiv2026

A cross-modal generative model for incomplete and degraded prostate MRI with multicentre clinical validation

Missing or degraded sequences can limit prostate multiparametric MRI. We developed MSCNet, a sequence-conditioned cross-modal generative framework for reconstructing unavailable contrasts and restoring degraded acquisitions. Across ten completion tasks, task-specific MSCNet achieved mean structural similarity of 0.818 versus 0.798 for the strongest task-matched comparators; matched-capacity analyses showed larger differences in lesion fidelity and boundary preservation. In a blinded 1,000-case reader study, overall image quality met the prespecified non-inferiority criterion for DWI, ADC and T2W completion, but not T1W. In a separate 200-case diagnostic assessment, AUCs for clinically significant cancer were 0.860 with acquired images, 0.841 with MSCNet and 0.797 with baseline-generated images. A locked 186-case three-hospital cohort supported multicentre transportability. These retrospective results support quality-controlled cross-modal reconstruction as an adjunct to acquired prostate MRI.

eess.IV↗
arXiv2026

Topological Feasibility Guarantees for Differentiable Predictive Control

Differentiable predictive control (DPC), a self-supervised learning approach for approximating explicit model predictive control (MPC) policies, offers significant computational advantages over online optimization-based MPC. However, feasibility guarantees, a core requirement for safe control, are currently provided either probabilistically or via online safety filters. The lack of rigorous feasibility guarantees for offline policy optimization remains an open problem. This paper establishes deterministic feasibility guarantees for DPC using a novel topological analysis of the induced reachable safe set, without requiring online safety filters. By exploiting the inherent model-based nature of DPC, in which differentiable system dynamics are embedded directly into the computational graph, we analyze the properties of the learned control policies and the corresponding system states from topological and geometric perspectives. Inspired by our theoretical analysis, we propose a novel self-supervised offline policy learning strategy that utilizes a proxy loss with Control Barrier Functions (CBFs). Crucially, these properties not only significantly improve policy training but also enable the derivation of strict, deterministic feasibility guarantees from a finite number of training samples. Extensive closed-loop simulations validate our theoretical findings, demonstrating that the empirical constraint violations monotonically decrease to zero as the training sample size increases. Ultimately, this work illustrates that DPC policy optimization yields formal safety certificates that are structurally unattainable with conventional black-box methods, e.g., reinforcement learning (RL) or supervised learning-based approximate MPC, thereby providing a new perspective on feasibility guarantees in learning-based control.

eess.SY↗
arXiv2026

Group Steering: Approaches Based on Power Moments

This paper considers the problem of steering a vast group of agents of which the dynamics are governed by a discrete-time asymptotically stable first-order linear system. The group of agents are characterized as a probability density function and an occupation measure respectively in the paper and two corresponding treatments are given. We propose to use the power moments to characterize the density function/occupation measure of the agents. A moment system representation of the original system is put forward for control and an empirical control scheme corresponding to it is proposed. By the designed control law, the moment sequence of the control at each time step is positive, which ensures the existence of the control for the moment system. We then realize the control as an analytic form of function by a convex optimization scheme of which the existence and uniqueness of the solution have been proved in our previous paper. The terminal density is proved to converge to the desired terminal one, which distinguishes the proposed distribution steering scheme from other existing ones. An error analysis of the terminal density from the specified one is also provided. For the problem where the group of agents is characterized as an occupation measure, the control for each agent is determined by drawing independent and identically-distributed(i.i.d) samples from the realized analytic function. Finally, simulation results validate our proposed algorithms.

math.OC↗
arXiv2026

Log-Insight: Automating Microservice Incident Diagnosis via Neuro-Symbolic Log Analysis

Diagnosing production incidents in large-scale microservice systems is time-critical for Site Reliability Engineers (SREs). A single 30-minute incident window in our deployment can generate over two million log lines--approximately 1.2 billion characters, far exceeding standard LLM context windows--making direct LLM-based Root Cause Analysis (RCA) infeasible. Existing approaches leave gaps: template-based parsers lack semantic anomaly reasoning, deep-learning detectors emit black-box binary signals, and LLM pipelines suffer context overflow and domain hallucination on raw telemetry. We present Log-Insight, an automated incident-diagnosis system deployed in production at Huawei. The core design principle automates the SRE's manual triage workflow: symbolic stages replicate the structured investigation a skilled SRE would perform--sampling, schema understanding, pattern clustering, and statistical anomaly ranking. This hands the LLM a compact, pre-ranked evidence dossier to synthesise into a hypothesis report. Our six-stage pipeline reduces millions of raw events by 1,000-7,000x while preserving statistically significant failure signals. Evaluated on 11 historical production incidents (110 runs, SRE-validated ground truth), Log-Insight achieves MRR = 0.790, returning the correct root cause within the top-3 hypotheses in over 90% of runs in under a minute of latency. We report systematic failure modes, active mitigations, and open research directions. The Forensic Evidence section--listing exact log templates and skew statistics--was consistently identified by operators as a key adoption factor, shifting the system's perceived role from opaque oracle to investigative assistant.

cs.IR↗
arXiv2026

Truncated Multidimensional Trigonometric Moment Problem: A Choice of Bases and the Unique Solution

In this paper, we resolve the Truncated Multidimensional Trigonometric Moment Problem (TMTMP) from a system and signal processing perspective, which serves as the foundation for the Multidimensional Rational Covariance Extension Problem (RCEP). While standard mathematical TMTMPs focus on the existence of atomic measure solutions, system identification requires analytic rational solutions with positive polynomial coefficients. To overcome the long-standing challenge of characterizing the positive feasible domain under general bases, we propose a novel choice of basis functions and a corresponding estimation scheme via convex optimization. We establish an explicit condition to guarantee the positiveness of the spectral estimate. Crucially, the map from the estimate parameters to the trigonometric moments is proved to be a diffeomorphism, ensuring the existence and uniqueness of the solution. Furthermore, we comprehensively prove the statistical properties of the estimator, including its consistency, (asymptotic) unbiasedness, convergence rate, and efficiency. The proposed framework is applied to a system identification task, where simulations validate its effectiveness.

math.OC↗
arXiv2026

Dual-Loop Control in DCVerse: Advancing Reliable Deployment of AI in Data Centers via Digital Twins

The growing scale and complexity of modern data centers present major challenges in balancing energy efficiency with outage risk. Although Deep Reinforcement Learning (DRL) shows strong potential for intelligent control, its deployment in mission-critical systems is limited by data scarcity and the lack of real-time pre-evaluation mechanisms. This paper introduces the Dual-Loop Control Framework (DLCF), a digital twin-based architecture designed to overcome these challenges. The framework comprises three core entities: the physical system, a digital twin, and a policy reservoir of diverse DRL agents. These components interact through a dual-loop mechanism involving real-time data acquisition, data assimilation, DRL policy training, pre-evaluation, and expert verification. Theoretical analysis shows how DLCF can improve sample efficiency, generalization, safety, and optimality. Leveraging DLCF, we implemented the DCVerse platform and validated it through case studies on a real-world data center cooling system. The evaluation shows that our approach achieves up to 4.09% energy savings over conventional control strategies without violating SLA requirements. Additionally, the framework improves policy interpretability and supports more trustworthy DRL deployment. This work provides a foundation for reliable AI-based control in data centers and points toward future extensions for holistic, system-wide optimization.

cs.AI↗
arXiv2026

Distribution Steering for Discrete-Time Uncertain Ensemble Systems

Ensemble systems appear frequently in many engineering applications and, as a result, they have become an important research topic in control theory. These systems are best characterized by the evolution of their underlying state distribution. Despite the work to date, few results exist dealing with the problem of directly modifying (i.e., ``steering'') the distribution of an ensemble system. In addition, in most existing results, the distribution of the states of an ensemble of discrete-time systems is assumed to be Gaussian. However, in case the system parameters are uncertain, it is not always realistic to assume that the distribution of the system follows a Gaussian distribution, thus complicating the solution of the overall problem. In this paper, we address the general distribution steering problem for first-order discrete-time ensemble systems, where the distributions of the system parameters and the states are arbitrary with finite first few moments. Linear system dynamics are considered using the method of power moments to transform the original infinite-dimensional problem into a finite-dimensional one. We also propose a control law for the ensuing moment system, which allows us to obtain the power moments of the desired control inputs. Finally, we solve the inverse problem to obtain the feasible control inputs from their corresponding power moments. We provide a numerical example to validate our theoretical developments.

math.OC↗
arXiv2025

LLM-Driven Stationarity-Aware Expert Demonstrations for Multi-Agent Reinforcement Learning in Mobile Systems

Multi-agent reinforcement learning (MARL) has been increasingly adopted in many real-world applications. While MARL enables decentralized deployment on resource-constrained edge devices, it suffers from severe non-stationarity due to the synchronous updates of agent policies. This non stationarity results in unstable training and poor policy con vergence, especially as the number of agents increases. In this paper, we propose RELED, a scalable MARL framework that integrates large language model (LLM)-driven expert demonstrations with autonomous agent exploration. RELED incorporates a Stationarity-Aware Expert Demonstration module, which leverages theoretical non-stationarity bounds to enhance the quality of LLM-generated expert trajectories, thus providing high reward and training-stable samples for each agent. Moreover, a Hybrid Expert-Agent Policy Optimization module adaptively balances each agent's learning from both expert-generated and agent-generated trajectories, accelerating policy convergence and improving generalization. Extensive experiments with real city networks based on OpenStreetMap demonstrate that RELED achieves superior performance compared to state-of-the-art MARL methods.

cs.LG↗
arXiv2025

CrossTrace: Efficient Cross-Thread and Cross-Service Span Correlation in Distributed Tracing for Microservices

Distributed tracing has become an essential technique for debugging and troubleshooting modern microservice-based applications, enabling software engineers to detect performance bottlenecks, identify failures, and gain insights into system behavior. However, implementing distributed tracing in large-scale applications remains challenging due to the need for extensive instrumentation. To reduce this burden, zero-code instrumentation solutions, such as those based on eBPF, have emerged, allowing span data to be collected without modifying application code. Despite this promise, span correlation, the process of establishing causal relationships between spans, remains a critical challenge in zero-code approaches. Existing solutions often rely on thread affinity, compromise system security by requiring the kernel integrity mode to be disabled, or incur significant computational overhead due to complex inference algorithms. This paper presents CrossTrace, a practical and efficient distributed tracing solution designed to support the debugging of microservice applications without requiring source code modifications. CrossTrace employs a greedy algorithm to infer intra-service span relationships from delay patterns, eliminating reliance on thread identifiers. For inter-service correlation, CrossTrace embeds span identifiers into TCP packet headers via eBPF, enabling secure and efficient correlation compromising system security policies. Evaluation results show that CrossTrace can correlate thousands of spans within seconds with over 90% accuracy, making it suitable for production deployment and valuable for microservice observability and diagnosis.

cs.NI↗
arXiv2025

General Distribution Steering: A Sub-Optimal Solution by Convex Optimization

General distribution steering is intrinsically an infinite-dimensional problem, when the continuous distributions to steer are arbitrary. We put forward a moment representation of the primal system for control in [42]. However, the system trajectory was a predetermined one without optimization towards a design criterion, which doesn't always ensure a most satisfactory solution. In this paper, we propose an optimization approach to the general distribution steering problem of the first-order discrete-time linear system, i.e., an optimal control law for the corresponding moment system. The domain of all feasible control inputs is non-convex and has a complex topology. We obtain a subset of it by minimizing a weighted sum of squared integral distances alongside the system trajectory. The feasible domain is then proved convex, and the optimal control problem can be treated as a convex optimization or by exhaustive search, based on the type of the cost function. Algorithms of steering for continuous and discrete distributions are then put forward respectively, by adopting a realization scheme of control inputs. We also provide an explicit advantage of our proposed algorithm by truncated power moments to the prevailing Gaussian Mixture Models. Experiments on different types of cost functions are given to validate the performance of our proposed algorithm. Since the moment system is a dimension-reduced counterpart of the primal system, we call this solution a sub-optimal one to the primal general distribution steering problem.

math.OC↗
arXiv2024

A Non-Gaussian Bayesian Filter Using Power and Generalized Logarithmic Moments

In this paper, we aim to propose a consistent non-Gaussian Bayesian filter of which the system state is a continuous function. The distributions of the true system states, and those of the system and observation noises, are only assumed Lebesgue integrable with no prior constraints on what function classes they fall within. This type of filter has significant merits in both theory and practice, which is able to ameliorate the curse of dimensionality for the particle filter, a popular non-Gaussian Bayesian filter of which the system state is parameterized by discrete particles and the corresponding weights. We first propose a new type of statistics, called the generalized logarithmic moments. Together with the power moments, they are used to form a density surrogate, parameterized as an analytic function, to approximate the true system state. The map from the parameters of the proposed density surrogate to both the power moments and the generalized logarithmic moments is proved to be a diffeomorphism, establishing the fact that there exists a unique density surrogate which satisfies both moment conditions. This diffeomorphism also allows us to use gradient methods to treat the convex optimization problem in determining the parameters. Last but not least, simulation results reveal the advantage of using both sets of moments for estimating mixtures of complicated types of functions. A robot localization simulation is also given, as an engineering application to validate the proposed filtering scheme.

math.OC↗
arXiv2023

A Multivariate Non-Gaussian Bayesian Filter Using Power Moments

In this paper, we extend our results on the univariate non-Gaussian Bayesian filter using power moments to the multivariate systems, which can be either linear or nonlinear. Doing this introduces several challenging problems, for example a positive parametrization of the density surrogate, which is not only a problem of filter design, but also one of the multiple dimensional Hamburger moment problem. We propose a parametrization of the density surrogate with the proofs to its existence, Positivstellensatz and uniqueness. Based on it, we analyze the errors of moments of the density estimates by the proposed density surrogate. A discussion on continuous and discrete treatments to the non-Gaussian Bayesian filtering problem is proposed to motivate the research on continuous parametrization of the system state. Simulation results on estimating different types of multivariate density functions are given to validate our proposed filter. To the best of our knowledge, the proposed filter is the first one implementing the multivariate Bayesian filter with the system state parameterized as a continuous function, which only requires the true states being Lebesgue integrable.

stat.ME↗
arXiv2023

General Discrete-Time Fokker-Planck Control by Power Moments

In this paper, we address the so-called general Fokker-Planck control problem for discrete-time first-order linear systems. Unlike conventional treatments, we don't assume the distributions of the system states to be Gaussian. Instead, we only assume the existence and finiteness of the first several order power moments of the distributions. It is proved in the literature that there doesn't exist a solution, which has a form of conventional feedback control, to this problem. We propose a moment representation of the system to turn the original problem into a finite-dimensional one. Then a novel feedback control term, which is a mixture of a feedback term and a Markovian transition kernel term is proposed to serve as the control input of the moment system. The states of the moment system are obtained by maximizing the smoothness of the state transition. The power moments of the transition kernels are obtained by a convex optimization problem, of which the solution is proved to exist and be unique. Then they are mapped back to the probability distributions. The control inputs to the original system are then obtained by sampling from the realized distributions. Simulation results are provided to validate our algorithm in treating the general discrete-time Fokker-Planck control problem.

math.OC↗
arXiv2023

Non-Gaussian Bayesian Filtering by Density Parametrization Using Power Moments

Non-Gaussian Bayesian filtering is a core problem in stochastic filtering. The difficulty of the problem lies in parameterizing the state estimates. However the existing methods are not able to treat it well. We propose to use power moments to obtain a parameterization. Unlike the existing parametric estimation methods, our proposed algorithm does not require prior knowledge about the state to be estimated, e.g. the number of modes and the feasible classes of function. Moreover, the proposed algorithm is not required to store massive parameters during filtering as the existing nonparametric Bayesian filters, e.g. the particle filter. The parameters of the proposed parametrization can also be determined by a convex optimization scheme with moments constraints, to which the solution is proved to exist and be unique. A necessary and sufficient condition for all the power moments of the density estimate to exist and be finite is provided. The errors of power moments are analyzed for the density estimate being either light-tailed or heavy-tailed. Error upper bounds of the density estimate for the one-step prediction are proposed. Simulation results on different types of density functions of the state are given, including the heavy-tailed densities, to validate the proposed algorithm.

stat.ME↗
arXiv2023

Density Steering by Power Moments

This paper considers the problem of steering an arbitrary initial probability density function to an arbitrary terminal one, where the system dynamics is governed by a first-order linear stochastic difference equation. It is a generalization of the conventional stochastic control problem where the uncertainty of the system state is usually characterized by a Gaussian distribution. We propose to use the power moments to turn the infinite-dimensional problem into a finite-dimensional one and to present an empirical control scheme. By the designed control law, the moment sequence of the controls at each time step is positive, which ensures the existence of the control for the moment system. We then realize the control at each time step as a function in analytic form by a convex optimization scheme, for which the existence and uniqueness of the solution have been proved in our previous paper. Two numerical examples are given to validate our proposed algorithm.

math.OC↗
arXiv2023

A Non-Classical Parameterization for Density Estimation Using Sample Moments

Probability density estimation is a core problem of statistics and signal processing. Moment methods are an important means of density estimation, but they are generally strongly dependent on the choice of feasible functions, which severely affects the performance. In this paper, we propose a non-classical parametrization for density estimation using sample moments, which does not require the choice of such functions. The parametrization is induced by the squared Hellinger distance, and the solution of it, which is proved to exist and be unique subject to a simple prior that does not depend on data, and can be obtained by convex optimization. Statistical properties of the density estimator, together with an asymptotic error upper bound are proposed for the estimator by power moments. Applications of the proposed density estimator in signal processing tasks are given. Simulation results validate the performance of the estimator by a comparison to several prevailing methods. To the best of our knowledge, the proposed estimator is the first one in the literature for which the power moments up to an arbitrary even order exactly match the sample moments, while the true density is not assumed to fall within specific function classes.

stat.ML↗
arXiv2023

Quantitative Error Analyses of Spectral Density Estimators Using Covariance Lags

Spectral density estimation is a core problem of system identification, which is an important research area of system control and signal processing. There have been numerous results on the design of spectral density estimators. However to our best knowledge, quantitative error analyses of the spectral density estimation have not been proposed yet. In real practice, there are two main factors which induce errors in the spectral density estimation, including the external additive noise and the limited number of samples. In this paper, which is a very preliminary version, we first consider a univariate spectral density estimator using covariance lags. The estimation task is performed by a convex optimization scheme, and the covariance lags of the estimated spectral density are exactly as desired, which makes it possible for quantitative error analyses such as to derive tight error upper bounds. We analyze the errors induced by the two factors and propose upper and lower bounds for the errors. Then the results of the univariate spectral estimator are generalized to the multivariate one.

math.OC↗
arXiv2021

Automatic Modulation Classification Using Involution Enabled Residual Networks

Automatic modulation classification (AMC) is of crucial importance for realizing wireless intelligence communications. Many deep learning based models especially convolution neural networks (CNNs) have been proposed for AMC. However, the computation cost is very high, which makes them inappropriate for beyond the fifth generation wireless communication networks that have stringent requirements on the classification accuracy and computing time. In order to tackle those challenges, a novel involution enabled AMC scheme is proposed by using the bottleneck structure of the residual networks. Involution is utilized instead of convolution to enhance the discrimination capability and expressiveness of the model by incorporating a self-attention mechanism. Simulation results demonstrate that our proposed scheme achieves superior classification performance and faster convergence speed comparing with other benchmark schemes.

eess.SP↗