arXiv subjects

Xiongtao Zhang

Publications and source records attributed to Xiongtao Zhang.

18 recordsLinked to original sources

arXiv2026

A multimodal large language model for evidence-based autism spectrum disorder screening

The clinical management of autism spectrum disorder (ASD) faces a bottleneck in early screening, mainly because trained specialists are scarce and conventional assessment tools are subjective. Here, we introduce ASDchat, a multimodal large language model designed for evidence-based ASD screening, which takes video, audio, and dialogue as input. ASDchat adopts a dual-branch architecture, where the decision branch generates screening probabilities and the evidence branch generates traceable, timestamped behavioral evidence aligned with standardized clinical criteria (ADOS-2). The model was trained and evaluated on a dataset of 1,035 participants from 27 sites in China, which covered typically developing (TD) children, children with ASD, and children with other disorders. For ASD versus TD, ASDchat reached an area under the receiver operating characteristic curve (AUC) of 0.953 $\pm$ 0.021. On 9 held-out sites that were not used for training, the mean AUC was 0.932. Furthermore, unsupervised clustering of the behavioral dimensions split the ASD cases into six subtypes with different phenotypic profiles, and ASDchat suggests an intervention for each subtype. ASDchat provides a feasible path for large-scale, evidence-based early ASD screening in clinical practice.

cs.CV↗

arXiv2026

FedUP: One-Shot Federated Unlearning via Centroid-Guided Plug-in Filters

Federated unlearning (FU) is critical for complying with legal mandates like the right to be forgotten in decentralized systems, yet current methods face a persistent dilemma between non-target knowledge loss and high request latency. To resolve these issues, we propose FedUP, a one-shot federated unlearning framework utilizing lightweight pluggable filters that act as a "knowledge funnel" to screen out target data while preserving original model performance. By freezing original model parameters and training filters at the server side using differentially private (DP)-protected class centroid samples, FedUP bypasses the need for multi-round client-server communication and complex retraining, reducing unlearning latency from minutes to mere seconds. Additionally, the framework's pluggable architecture ensures inherent reversibility, enabling the seamless restoration of forgotten knowledge by simply removing the filters. Extensive experiments on diverse image and text tasks demonstrate that FedUP effectively reduces non-target knowledge loss and achieves superior unlearning precision and efficiency across various scenarios. Code is available at: https://github.com/suows/FedUP-code.

cs.LG↗

arXiv2026

Make LLM Learn to Synthesize from Streaming Experiences through Feedback

Large language models (LLMs) have been widely adopted for synthetic data generation, significantly reducing annotation costs. However, most existing studies treat synthesis as a set of isolated tasks and overlook a more fundamental question: whether a model can learn to synthesize by accumulating experience from past tasks and transferring it to future ones. In this work, we introduce StreamSynth, a new setting in which synthesis tasks arrive sequentially and experience from historical tasks provides informative signals for future synthesis. To address this setting, we propose SynLearner, a general framework that enables synthesis models to acquire reusable synthesis experience over a task stream. Instead of generating data independently for each task, SynLearner encourages the model to explore diverse synthesis patterns, learn from feedback, and balance sample quality with set-level diversity as tasks evolve. Extensive experiments across multiple benchmarks show that SynLearner effectively leverages experience from earlier tasks to improve synthesis performance on later ones, exhibiting consistent cross-task transferability. These findings provide evidence for the feasibility of StreamSynth and highlight synthetic data generation as an experience-driven process that can benefit from task streams.

cs.AI↗

arXiv2025

Synchronization of second-order Kuramoto model with frustration on strongly connected digraph

We study the emergent behavior of a second-order Kuramoto-type model with frustration effect on a strongly connected digraph. The main challenge arises from the lack of symmetry in this system, which renders standard approaches for symmetric models, such as the gradient-flow method and classical $\ell^p$ or $\ell^\infty$-type energy estimates, ineffective. To address these difficulties, our primary contribution is the development of time-dependent weighted $\ell^1$-type energy estimates to establish the hypo-coercivity of the frequency diameter. Specifically, we construct novel energy functions incorporating convex combinations of phases, frequencies, accelerations, and jerks, which are shown to be dissipative and capable of bounding both phase and frequency diameters. This framework enables us to demonstrate the emergence of frequency synchronization with an exponential convergence rate.

math.DS↗

arXiv2025

Multi-task Federated Learning with Encoder-Decoder Structure: Enabling Collaborative Learning Across Different Tasks

Federated learning has been extensively studied and applied due to its ability to ensure data security in distributed environments while building better models. However, clients participating in federated learning still face limitations, as clients with different structures or tasks cannot participate in learning together. In view of this, constructing a federated learning framework that allows collaboration between clients with different model structures and performing different tasks, enabling them to share valuable knowledge to enhance model efficiency, holds significant practical implications for the widespread application of federated learning. To achieve this goal, we propose a multi-task federated learning with encoder-decoder structure (M-Fed). Specifically, given the widespread adoption of the encoder-decoder architecture in current models, we leverage this structure to share intra-task knowledge through traditional federated learning methods and extract general knowledge from the encoder to achieve cross-task knowledge sharing. The training process is similar to traditional federated learning, and we incorporate local decoder and global decoder information into the loss function. The local decoder iteratively updates and gradually approaches the global decoder until sufficient cross-task knowledge sharing is achieved. Our method is lightweight and modular, demonstrating innovation compared to previous research. It enables clients performing different tasks to share general knowledge while maintaining the efficiency of traditional federated learning systems. We conducted experiments on two widely used benchmark datasets to verify the feasibility of M-Fed and compared it with traditional methods. The experimental results demonstrate the effectiveness of M-Fed in multi-task federated learning.

cs.LG↗

arXiv2025

On the exponential synchronization for the asymmetric second-order Kuramoto model

In this paper, we study the synchronization problem of nonuniform second-order Kuramoto model with homogeneous dampings and frustration effects on an asymmetric network. More precisely, we focus on the second order model defined on an asymmetric graph with depth no greater than two and present theories on the complete frequency synchronization. Due to the absence of the gradient flow structure, we develop novel energy functions to control the diameters of phase and frequency respectively, which allows us to construct first-order Gronwall-type inequalities. This eventually gives rise to the exponential convergence to the synchronized state in a regime in terms of large coupling strength, small inertia and frustration.

math.DS↗

arXiv2025

FedHPD: Heterogeneous Federated Reinforcement Learning via Policy Distillation

Federated Reinforcement Learning (FedRL) improves sample efficiency while preserving privacy; however, most existing studies assume homogeneous agents, limiting its applicability in real-world scenarios. This paper investigates FedRL in black-box settings with heterogeneous agents, where each agent employs distinct policy networks and training configurations without disclosing their internal details. Knowledge Distillation (KD) is a promising method for facilitating knowledge sharing among heterogeneous models, but it faces challenges related to the scarcity of public datasets and limitations in knowledge representation when applied to FedRL. To address these challenges, we propose Federated Heterogeneous Policy Distillation (FedHPD), which solves the problem of heterogeneous FedRL by utilizing action probability distributions as a medium for knowledge sharing. We provide a theoretical analysis of FedHPD's convergence under standard assumptions. Extensive experiments corroborate that FedHPD shows significant improvements across various reinforcement learning benchmark tasks, further validating our theoretical findings. Moreover, additional experiments demonstrate that FedHPD operates effectively without the need for an elaborate selection of public datasets.

cs.LG↗

arXiv2024

Exponential synchronization of the Kuramoto model with inertia and frustration under locally coupled network

We study the collective synchronized behavior of the Kuramoto model with inertia and frustration effects on a connected and symmetric network. We aim to establish sufficient frameworks for achieving complete frequency synchronization, taking into account initial configuration, small inertia and frustration, and large coupling strength. More precisely, we first demonstrate that the phase diameter will be uniformly bounded by a small value after a finite time. Then we prove that the frequency diameter exhibits exponential decay to zero. Our approach relies on a careful construction of energy functionals, which effectively control the dissipation of phase and frequency diameters.

math.DS↗

arXiv2024

Emergent dynamics of the inertial Kuramoto model with frustration on a locally coupled graph

We study the synchronized behavior of the inertial Kuramoto oscillators with frustration effect under a symmetric and connected network. Due to the lack of second-order gradient flow structure and singularity of second-order derivative of diameter, we shift to construct convex combinations of oscillators and related new energy functions that can control the phase and frequency diameters. Under sufficient frameworks on initial data and system parameters, we derive first-order dissipative differential inequalities of constructed energy functions. This eventually gives rise to the emergence of frequency synchronization exponentially fast.

math.DS↗

arXiv2023

Fuzzy Knowledge Distillation from High-Order TSK to Low-Order TSK

High-order Takagi-Sugeno-Kang (TSK) fuzzy classifiers possess powerful classification performance yet have fewer fuzzy rules, but always be impaired by its exponential growth training time and poorer interpretability owing to High-order polynomial used in consequent part of fuzzy rule, while Low-order TSK fuzzy classifiers run quickly with high interpretability, however they usually require more fuzzy rules and perform relatively not very well. Address this issue, a novel TSK fuzzy classifier embeded with knowledge distillation in deep learning called HTSK-LLM-DKD is proposed in this study. HTSK-LLM-DKD achieves the following distinctive characteristics: 1) It takes High-order TSK classifier as teacher model and Low-order TSK fuzzy classifier as student model, and leverages the proposed LLM-DKD (Least Learning Machine based Decoupling Knowledge Distillation) to distill the fuzzy dark knowledge from High-order TSK fuzzy classifier to Low-order TSK fuzzy classifier, which resulting in Low-order TSK fuzzy classifier endowed with enhanced performance surpassing or at least comparable to High-order TSK classifier, as well as high interpretability; specifically 2) The Negative Euclidean distance between the output of teacher model and each class is employed to obtain the teacher logits, and then it compute teacher/student soft labels by the softmax function with distillating temperature parameter; 3) By reformulating the Kullback-Leibler divergence, it decouples fuzzy dark knowledge into target class knowledge and non-target class knowledge, and transfers them to student model. The advantages of HTSK-LLM-DKD are verified on the benchmarking UCI datasets and a real dataset Cleveland heart disease, in terms of classification performance and model interpretability.

cs.LG↗

arXiv2023

TcGAN: Semantic-Aware and Structure-Preserved GANs with Individual Vision Transformer for Fast Arbitrary One-Shot Image Generation

One-shot image generation (OSG) with generative adversarial networks that learn from the internal patches of a given image has attracted world wide attention. In recent studies, scholars have primarily focused on extracting features of images from probabilistically distributed inputs with pure convolutional neural networks (CNNs). However, it is quite difficult for CNNs with limited receptive domain to extract and maintain the global structural information. Therefore, in this paper, we propose a novel structure-preserved method TcGAN with individual vision transformer to overcome the shortcomings of the existing one-shot image generation methods. Specifically, TcGAN preserves global structure of an image during training to be compatible with local details while maintaining the integrity of semantic-aware information by exploiting the powerful long-range dependencies modeling capability of the transformer. We also propose a new scaling formula having scale-invariance during the calculation period, which effectively improves the generated image quality of the OSG model on image super-resolution tasks. We present the design of the TcGAN converter framework, comprehensive experimental as well as ablation studies demonstrating the ability of TcGAN to achieve arbitrary image generation with the fastest running time. Lastly, TcGAN achieves the most excellent performance in terms of applying it to other image processing tasks, e.g., super-resolution as well as image harmonization, the results further prove its superiority.

cs.CV↗

arXiv2022

Mask-Guided Image Person Removal with Data Synthesis

As a special case of common object removal, image person removal is playing an increasingly important role in social media and criminal investigation domains. Due to the integrity of person area and the complexity of human posture, person removal has its own dilemmas. In this paper, we propose a novel idea to tackle these problems from the perspective of data synthesis. Concerning the lack of dedicated dataset for image person removal, two dataset production methods are proposed to automatically generate images, masks and ground truths respectively. Then, a learning framework similar to local image degradation is proposed so that the masks can be used to guide the feature extraction process and more texture information can be gathered for final prediction. A coarse-to-fine training strategy is further applied to refine the details. The data synthesis and learning framework combine well with each other. Experimental results verify the effectiveness of our method quantitatively and qualitatively, and the trained network proves to have good generalization ability either on real or synthetic images.

cs.CV↗

arXiv2021

Compressible Navier-Stokes equation with BV initial data: Part II. Global stability

In previous work \cite{W-Y-Z-local}, we studied the local well-posedness of weak solution to the 1-D full compressible Navier-Stokes equation with initial data of small total variation. Specifically, the local existence, the regularity, and the uniqueness in certain function space of the weak solution have been established. The basis for the previous study is the precise construction of fundamental solution for heat equation with BV conductivity. In this paper, we continue to investigate the global stability and the time asymptotic behavior of the weak solution. The main step is to construct the ``effective Green's function'', which is the combination of the heat kernel with BV coefficient in short time and the Green's function around constant state in long time. The former one captures the quasi-linear nature of the system, while the latter one respects the dissipative structure. Then the weak solution is written into an integral system in terms of this ``effective Green's function'', and the time asymptotic behavior is established based on a priori estimate.

math.AP↗

arXiv2021

On the Stochastic Singular Cucker--Smale Model: Well-Posedness, Collision-Avoidance and Flocking

We study the Cucker--Smale (C-S) flocking systems involving both singularity and noise. We first show the local strong well-posedness for the stochastic singular C-S systems before the first collision time, which is a well defined stopping time. Then, for communication with higher order singularity at origin (corresponding to $α\ge1$ in the case of $ψ(r)=r^{-α}$), we establish the global well-posedness by showing the collision-avoidance in finite time, provided that there is no initial collisions and the initial velocities have finite moment of any positive order. Finally, we study the large time behavior of the solution when $ψ$ is of zero lower bound, and provide the emergence of conditional flocking or unconditional flocking in the mean sense, for constant and square integrable intensity respectively.

math.DS↗

arXiv2021

Emergence of synchronization in Kuramoto model with general digraph

In this paper, we study the complete synchronization of the Kuramoto model with general network containing a spanning tree, when the initial phases are distributed in an open half circle. As lack of uniform coercivity in general digraph, in order to capture the dissipation structure on a general network, we apply the node decomposition criteria in \cite{H-L-Z20} to yield a hierarchical structure, which leads to the hypo-coercivity. This drives the phase diameter into a small region after finite time in a large coupling regime, and the uniform boundedness of the diameter eventually leads to the emergence of exponentially fast synchronization.

math.DS↗

arXiv2020

One dimensional singular Cucker-Smale model: uniform-in-time mean-field limit and contractivity

We analyze the one dimensional Cucker-Smale (in short CS) model with a weak singular communication weight $ψ(x) = |x|^{-β}$ with $β\in (0,1)$. We first establish a global-in-time existence of measure-valued solutions to the kinetic CS equation. For this, we use a proper change of variable to reformulate the particle CS model as a first-order particle system and provide the uniform-in-time stability for that particle system. We then extend this stability estimate for the singular CS particle system. By using that stability estimate, we construct the measure-valued solutions to the kinetic CS equation globally in time. Moreover, as a direct application of the uniform-in-time stability estimate, we show the quantitative uniform-in-time mean-field limit from the particle system to that kinetic CS equation in $p$-Wasserstein distance with $p \in [1,\infty]$. Our result gives the uniqueness of measure-valued solution in the sense of mean-field limits, i.e., the measure-valued solutions, approximated by the empirical measures associated to the particle system, uniquely exist. Similar results for the first-order model also follow as a by-product. We also reformulate the continuity-type equation, which is derived from the first-order model, as an integro-differential equation by employing the pseudo-inverse of the accumulative particle distribution. By making use of a modified $p$-Wasserstein distance, we provide the contractivity estimate for absolutely continuous solutions of the continuum equation.

math.AP↗

arXiv2019

Complete classification of the asymptotical behavior for singular C-S Model on the real line

In this paper, we study the singular Cucker-Smale (C-S) model on the real line. For long range case, i.e. $β<1$, we prove the uniqueness of the solution in the sense of Definition 2.1 and the unconditional flocking emergence. Moreover, the sufficient and necessary condition for collision and sticking phenomenon will be provided. For short range case, i.e. $β>1$, we construct the uniform-in-time lower bound of the relative distance between particles and provide the sufficient and necessary condition for the emergence of multi-cluster formation. For critical case, i.e. $β=1$, we show the uniform lower bound of the relative distance and unconditional flocking emergence. These results provide a complete classification of the collective behavior for C-S model on the real line.

math.DS↗

arXiv2019

Emergent behaviors of the discrete-time Kuramoto model for generic initial configuration

In this paper, we will study the emergent dynamics of the discrete Kuramoto model for generic initial data. This is an extension of the previous work S.-Y. Ha et al. (2019), in which the initial configurations are supposed to be within a half circle. More precisely, we will provide the theory of discrete gradient flow which can be applied to general Euler iteration scheme. Therefore, as a direct application, we conclude the emergence of synchronization of discrete Kuramoto model. Moreover, we obtain for small mesh size that, the synchronization will occur exponentially fast for initial data in A_1 (see definition in (4.1)).

math.DS↗