arXiv subjects

Wentao Fan

Publications and source records attributed to Wentao Fan.

13 recordsLinked to original sources

arXiv2026

How Do Optical Flow and Textual Prompts Collaborate to Assist in Audio-Visual Semantic Segmentation?

Audio-visual semantic segmentation (AVSS) represents an extension of the audio-visual segmentation (AVS) task, necessitating a semantic understanding of audio-visual scenes beyond merely identifying sound-emitting objects at the visual pixel level. Contrary to a previous methodology, by decomposing the AVSS task into two discrete subtasks by initially providing a prompted segmentation mask to facilitate subsequent semantic analysis, our approach innovates on this foundational strategy. We introduce a novel collaborative framework, \textit{S}tepping \textit{S}tone \textit{P}lus (SSP), which integrates optical flow and textual prompts to assist the segmentation process. In scenarios where sound sources frequently coexist with moving objects, our pre-mask technique leverages optical flow to capture motion dynamics, providing essential temporal context for precise segmentation. To address the challenge posed by stationary sound-emitting objects, such as alarm clocks, SSP incorporates two specific textual prompts: one identifies the category of the sound-emitting object, and the other provides a broader description of the scene. Additionally, we implement a visual-textual alignment module (VTA) to facilitate cross-modal integration, delivering more coherent and contextually relevant semantic interpretations. Our training regimen involves a post-mask technique aimed at compelling the model to learn the diagram of the optical flow. Experimental results demonstrate that SSP outperforms existing AVS methods, delivering efficient and precise segmentation results.

cs.CV↗

arXiv2025

Enhancing Rotation-Invariant 3D Learning with Global Pose Awareness and Attention Mechanisms

Recent advances in rotation-invariant (RI) learning for 3D point clouds typically replace raw coordinates with handcrafted RI features to ensure robustness under arbitrary rotations. However, these approaches often suffer from the loss of global pose information, making them incapable of distinguishing geometrically similar but spatially distinct structures. We identify that this limitation stems from the restricted receptive field in existing RI methods, leading to Wing-tip feature collapse, a failure to differentiate symmetric components (e.g., left and right airplane wings) due to indistinguishable local geometries. To overcome this challenge, we introduce the Shadow-informed Pose Feature (SiPF), which augments local RI descriptors with a globally consistent reference point (referred to as the 'shadow') derived from a learned shared rotation. This mechanism enables the model to preserve global pose awareness while maintaining rotation invariance. We further propose Rotation-invariant Attention Convolution (RIAttnConv), an attention-based operator that integrates SiPFs into the feature aggregation process, thereby enhancing the model's capacity to distinguish structurally similar components. Additionally, we design a task-adaptive shadow locating module based on the Bingham distribution over unit quaternions, which dynamically learns the optimal global rotation for constructing consistent shadows. Extensive experiments on 3D classification and part segmentation benchmarks demonstrate that our approach substantially outperforms existing RI methods, particularly in tasks requiring fine-grained spatial discrimination under arbitrary rotations.

cs.CV↗

arXiv2025

Ergodicity of asymmetric lemon Billiards

We show ergodicity of (asymmetric) lemon billiards, billiard tables that are the intersection of two circles of which one contains the centers of both. These do not satisfy the Wojtkowski criteria for hyperbolicity, but we establish \emph{uniform} expansion of vectors in an invariant cone family and alignment of singularity curves. Both of these are difficult, and the approach to the latter seems new. Together, these properties imply ergodicity.

math.DS↗

arXiv2025

Generalized Time-Coarse Graining via an Operator Cumulant Expansion

We introduce a general framework for deriving effective dynamics from arbitrary time-dependent generators, based on a systematic operator cumulant expansion. Unlike traditional approaches, which typically assume periodic or adiabatic driving, our method applies to systems with general time dependencies and is compatible with any dynamics generated by a linear operator -- Hamiltonian or not, quantum or classical, open or closed. This enables modeling of systems exhibiting strong modulation, dissipation, or non-adiabatic effects. Our approach unifies Hamiltonian techniques such as Lie-transform Perturbation Theory (LPT) with averaging-based methods like Time-Coarse Graining (TCG), revealing their structural equivalence through the lens of generalized cumulants. It also clarifies how non-Hamiltonian terms naturally emerge from averaging procedures, even in closed systems. We illustrate the power and flexibility of the method by analyzing a damped, parametrically driven Kapitza pendulum, a system beyond the reach of standard tools, demonstrating how accurate effective equations can be derived across a wide range of regimes.

math-ph↗

arXiv2025

Long-time soliton dynamics via a coarse-grained space-time method

We investigate the long-time dynamics of the Sine-Gordon (SG) model under a class of perturbations whose quantum field theoretic analog - via bosonization - corresponds to the massive Schwinger model describing 1+1D relativistic QED of Dirac fermions. Classical SG solutions offer critical insight into non-perturbative effects in this quantum theory, but capturing their long-time behavior poses significant numerical challenges. To address this, we extend a coarse-graining method to spacetime using a dual-mesh construction based on the Minkowski-metric. We first validate the approach against the well-studied variant of the SG model describing magnetic fluxon dynamics in Josephson transmission lines (JTLs), where analytical and numerical benchmarks exist. We then apply the method to the Schwinger-inspired SG model and uncover long-lived bound states - "Schwinger atoms" - in which a soliton is trapped by a fixed central charge. In certain regimes, the system exhibits limit cycles that give rise to positronium-like states of oppositely charged solitons, while in others such formation is suppressed. Accessing such long-time solutions requires a rigorous implementation of outgoing boundary conditions on a finite computational domain that provide radiative dissipation to allow relaxation toward states that exist only in an infinite domain. Here we provide such a construction. Our results also suggest the possibility of analog quantum simulation of relativistic quantum field theories with JTLs. These results demonstrate the utility of spatio-temporal coarse-graining methodology for probing non-perturbative structure formation in non-linear field theories.

quant-ph↗

arXiv2024

Accelerating AIGC Services with Latent Action Diffusion Scheduling in Edge Networks

Artificial Intelligence Generated Content (AIGC) has gained significant popularity for creating diverse content. Current AIGC models primarily focus on content quality within a centralized framework, resulting in a high service delay and negative user experiences. However, not only does the workload of an AIGC task depend on the AIGC model's complexity rather than the amount of data, but the large model and its multi-layer encoder structure also result in a huge demand for computational and memory resources. These unique characteristics pose new challenges in its modeling, deployment, and scheduling at edge networks. Thus, we model an offloading problem among edges for providing real AIGC services and propose LAD-TS, a novel Latent Action Diffusion-based Task Scheduling method that orchestrates multiple edge servers for expedited AIGC services. The LAD-TS generates a near-optimal offloading decision by leveraging the diffusion model's conditional generation capability and the reinforcement learning's environment interaction ability, thereby minimizing the service delays under multiple resource constraints. Meanwhile, a latent action diffusion strategy is designed to guide decision generation by utilizing historical action probability, enabling rapid achievement of near-optimal decisions. Furthermore, we develop DEdgeAI, a prototype edge system with a refined AIGC model deployment to implement and evaluate our LAD-TS method. DEdgeAI provides a real AIGC service for users, demonstrating up to 29.18% shorter service delays than the current five representative AIGC platforms. We release our open-source code at https://github.com/ChangfuXu/DEdgeAI/.

cs.LG↗

arXiv2024

Model Order Reduction for Open Quantum Systems Based on Measurement-adapted Time-coarse Graining

Model order reduction encompasses mathematical techniques aimed at reducing the complexity of mathematical models in simulations while retaining the essential characteristics and behaviors of the original model. This is particularly useful in the context of large-scale dynamical systems, which can be computationally expensive to analyze and simulate. Here, we present a model order reduction technique to reduce the time complexity of open quantum systems, grounded in the principle of measurement-adapted coarse-graining. This method, governed by a coarse-graining time scale $\tau$ and the spectral band center $\omega_0$, organizes corrections to the lowest-order model which aligns with the RWA Hamiltonian in certain limits, and rigorously justifies the resulting effective quantum master equation (EQME). The focus on calculating to a high degree of accuracy only what can be resolved by the measurement introduces a principled regularization procedure to address singularities and generates low-stiffness models suitable for efficient long-time integration. Furthermore, the availability of the analytical form of the EQME parameters significantly boosts the interpretive capabilities of the method. As a demonstration, we derive the fourth-order EQME for a challenging problem related to the dynamics of a superconducting qubit under high-power dispersive readout in the presence of a continuum of dissipative waveguide modes. This derivation shows that the lowest-order terms align with previous results, while higher-order corrections suggest new phenomena.

quant-ph↗

arXiv2024

Leveraging CORAL-Correlation Consistency Network for Semi-Supervised Left Atrium MRI Segmentation

Semi-supervised learning (SSL) has been widely used to learn from both a few labeled images and many unlabeled images to overcome the scarcity of labeled samples in medical image segmentation. Most current SSL-based segmentation methods use pixel values directly to identify similar features in labeled and unlabeled data. They usually fail to accurately capture the intricate attachment structures in the left atrium, such as the areas of inconsistent density or exhibit outward curvatures, adding to the complexity of the task. In this paper, we delve into this issue and introduce an effective solution, CORAL(Correlation-Aligned)-Correlation Consistency Network (CORN), to capture the global structure shape and local details of Left Atrium. Diverging from previous methods focused on each local pixel value, the CORAL-Correlation Consistency Module (CCM) in the CORN leverages second-order statistical information to capture global structural features by minimizing the distribution discrepancy between labeled and unlabeled samples in feature space. Yet, direct construction of features from unlabeled data frequently results in ``Sample Selection Bias'', leading to flawed supervision. We thus further propose the Dynamic Feature Pool (DFP) for the CCM, which utilizes a confidence-based filtering strategy to remove incorrectly selected features and regularize both teacher and student models by constraining the similarity matrix to be consistent. Extensive experiments on the Left Atrium dataset have shown that the proposed CORN outperforms previous state-of-the-art semi-supervised learning methods.

cs.CV↗

arXiv2024

Systematic time-coarse graining for driven quantum systems

Many recent advancements in quantum computing leverage strong drives on nonlinear systems for state preparation, signal amplification, or gate operation. However, the interplay within such strongly driven system introduces multi-scale dynamics that affects the long-time behavior of the system in non-trivial ways that are very difficult to model. Therefore, the analysis of these systems often relies on effective Hamiltonian models that introduce additional nonlinear processes which approximate the long-time dynamics so that highly oscillatory terms may be ignored. However, the removal of such high frequency transitions can only be performed rigorously within a systematic framework of time-coarse graining, which is a fundamentally irreversible operation. This implies that standard approaches with unitary effective models cannot accurately capture the long-time behavior of strongly driven nonlinear quantum systems in general. We introduce a systematic perturbation theory for obtaining the complete non-unitary effective model of the time-coarse grained (TCG) dynamics of a driven quantum system to any order in the coupling strengths. We derive a closed-form analytical formula for both unitary and non-unitary contributions, in the form of an effective Hamiltonian and non-unitary (pseudo-)dissipators. Remarkably, even though the effective theory presumes unitary time evolution at the microscopic level, the time-coarse grained dynamics is found to follow a non-unitary time evolution in general. This occurs even when there is no open heat reservoir for the system to become entangled with or dissipate into. We demonstrate the effectiveness of the new method using several typical models of driven nonlinear systems in superconducting circuits, and show that it generalizes and improves on existing methods by providing more accurate results and explaining phenomena that have not been accounted for.

quant-ph↗

arXiv2023

DEC-QED: A flux-based 3D electrodynamic modeling approach to superconducting circuits and materials

Modeling the behavior of superconducting electronic circuits containing Josephson junctions is crucial for the design of superconducting information processors and devices. In this paper, we introduce DEC-QED, a computational approach for modeling the electrodynamics of superconducting electronic circuits containing Josephson junctions in arbitrary three-dimensional electromagnetic environments. DEC-QED captures the non-linear response and induced currents in BCS superconductors and accurately captures phenomena such as the Meissner effect, flux quantization and Josephson effects. Using a spatial coarse-graining formulation based on Discrete Exterior Calculus (DEC), DEC-QED can accurately simulate transient and long-time dynamics in superconductors. The expression of the entire electrodynamic problem in terms of the gauge-invariant flux field and charges makes the resulting classical field theory suitable for second quantization.

quant-ph↗

arXiv2017

Motives and periods in Bianchi IX gravity models

We show that, when considering the anisotropic scaling factors and their derivatives as affine variables, the coefficients of the heat kernel expansion of the Dirac-Laplacian on $SU(2)$ Bianchi IX metrics are algebro-geometric periods of motives of complements in affine spaces of unions of quadrics and hyperplanes. We show that the motives are mixed Tate and we provide an explicit computation of their Grothendieck classes.

math-ph↗

arXiv2015

Modular forms in the spectral action of Bianchi IX gravitational instantons

Following our result on rationality of the spectral action for Bianchi type-IX cosmological models, which suggests the existence of a rich arithmetic structure in the action, we study the arithmetic and modular properties of the action for an especially interesting family of metrics, namely $SU(2)$-invariant Bianchi IX gravitational instantons. A variant of the previous rationality result is first presented for a time-dependent conformal perturbation of the triaxial Bianchi IX metric which serves as a general form of Bianchi IX gravitational instantons. By exploiting a parametrization of the gravitational instantons in terms of theta functions with characteristics, we show that each Seeley-de Witt coefficient $\tilde a_{2n}$ appearing in the expansion of their spectral action gives rise to vector-valued and ordinary modular functions of weight 2. We investigate the modular properties of the first three terms $\tilde a_0$, $\tilde a_2$ and $\tilde a_4$ by preforming direct calculations. This guides us to provide spectral arguments for general terms $\tilde a_{2n}$, which are based on isospectrality of the involved Dirac operators. Special attention is paid to the relation of the new modular functions with well-known modular forms by studying explicitly two different cases of metrics with pairs of rational parameters that belong to two different general families. In the first case, it is shown that by running a summation over a finite orbit of the rational parameters, up to multiplication by the cusp form $Δ$ of weight 12, each term $\tilde a_{2n}$ lands in the 1-dimensional linear space of modular forms of weight 14. In the second case, it is proved that, after a summation over a finite orbit of the parameters and multiplication by $G_4^4$, where $G_4$ is the Eisenstein series of weight 4, each $\tilde a_{2n}$ lands in the 1-dimensional linear space of cusp forms of weight 18.

math.DG↗

arXiv2015

Spectral Action for Bianchi Type-IX Cosmological Models

A rationality result previously proved for Robertson-Walker metrics is extended to a homogeneous anisotropic cosmological model, namely the Bianchi type-IX minisuperspace. It is shown that the Seeley-de Witt coefficients appearing in the expansion of the spectral action for the Bianchi type-IX geometry are expressed in terms of polynomials with rational coefficients in the cosmic evolution factors $w_1(t), w_2(t), w_3(t),$ and their higher derivates with respect to time. We begin with the computation of the Dirac operator of this geometry and calculate the coefficients $a_0, a_2, a_4$ of the spectral action by using heat kernel methods and parametric pseudodifferential calculus. An efficient method is devised for computing the Seeley-de Witt coefficients of a geometry by making use of Wodzicki's noncommutative residue, and it is confirmed that the method checks out for the cosmological model studied in this article. The advantages of the new method are discussed, which combined with symmetries of the Bianchi type-IX metric, yield an elegant proof of the rationality result.

hep-th↗