arXiv subjects

Haolin Li

Publications and source records attributed to Haolin Li.

At least 37 records · Page 2Linked to original sources

arXiv2025

Exploring Semantic-constrained Adversarial Example with Instruction Uncertainty Reduction

Recently, semantically constrained adversarial examples (SemanticAE), which are directly generated from natural language instructions, have become a promising avenue for future research due to their flexible attacking forms. To generate SemanticAEs, current methods fall short of satisfactory attacking ability as the key underlying factors of semantic uncertainty in human instructions, such as referring diversity, descriptive incompleteness, and boundary ambiguity, have not been fully investigated. To tackle the issues, this paper develops a multi-dimensional instruction uncertainty reduction (InSUR) framework to generate more satisfactory SemanticAE, i.e., transferable, adaptive, and effective. Specifically, in the dimension of the sampling method, we propose the residual-driven attacking direction stabilization to alleviate the unstable adversarial optimization caused by the diversity of language references. By coarsely predicting the language-guided sampling process, the optimization process will be stabilized by the designed ResAdv-DDIM sampler, therefore releasing the transferable and robust adversarial capability of multi-step diffusion models. In task modeling, we propose the context-encoded attacking scenario constraint to supplement the missing knowledge from incomplete human instructions. Guidance masking and renderer integration are proposed to regulate the constraints of 2D/3D SemanticAE, activating stronger scenario-adapted attacks. Moreover, in the dimension of generator evaluation, we propose the semantic-abstracted attacking evaluation enhancement by clarifying the evaluation boundary, facilitating the development of more effective SemanticAE generators. Extensive experiments demonstrate the superiority of the transfer attack performance of InSUR. Moreover, we realize the reference-free generation of semantically constrained 3D adversarial examples for the first time.

cs.AI↗

arXiv2025

FedGTEA: Federated Class-Incremental Learning with Gaussian Task Embedding and Alignment

We introduce a novel framework for Federated Class Incremental Learning, called Federated Gaussian Task Embedding and Alignment (FedGTEA). FedGTEA is designed to capture task-specific knowledge and model uncertainty in a scalable and communication-efficient manner. At the client side, the Cardinality-Agnostic Task Encoder (CATE) produces Gaussian-distributed task embeddings that encode task knowledge, address statistical heterogeneity, and quantify data uncertainty. Importantly, CATE maintains a fixed parameter size regardless of the number of tasks, which ensures scalability across long task sequences. On the server side, FedGTEA utilizes the 2-Wasserstein distance to measure inter-task gaps between Gaussian embeddings. We formulate the Wasserstein loss to enforce inter-task separation. This probabilistic formulation not only enhances representation learning but also preserves task-level privacy by avoiding the direct transmission of latent embeddings, aligning with the privacy constraints in federated learning. Extensive empirical evaluations on popular datasets demonstrate that FedGTEA achieves superior classification performance and significantly mitigates forgetting, consistently outperforming strong existing baselines.

cs.LG↗

arXiv2025

Cross-Resolution SAR Target Detection Using Structural Hierarchy Adaptation and Reliable Adjacency Alignment

In recent years, continuous improvements in SAR resolution have significantly benefited applications such as urban monitoring and target detection. However, the improvement in resolution leads to increased discrepancies in scattering characteristics, posing challenges to the generalization ability of target detection models. While domain adaptation technology is a potential solution, the inevitable discrepancies caused by resolution differences often lead to blind feature adaptation and unreliable semantic propagation, ultimately degrading the domain adaptation performance. To address these challenges, this paper proposes a novel SAR target detection method (termed CR-Net), that incorporates structure priors and evidential learning theory into the detection model, enabling reliable domain adaptation for cross-resolution detection. To be specific, CR-Net integrates Structure-induced Hierarchical Feature Adaptation (SHFA) and Reliable Structural Adjacency Alignment (RSAA). SHFA module is introduced to establish structural correlations between targets and achieve structure-aware feature adaptation, thereby enhancing the interpretability of the feature adaptation process. Afterwards, the RSAA module is proposed to enhance reliable semantic alignment, by leveraging the secure adjacency set to transfer valuable discriminative knowledge from the source domain to the target domain. This further improves the discriminability of the detection model in the target domain. Based on experimental results from different-resolution datasets,the proposed CR-Net significantly enhances cross-resolution adaptation by preserving intra-domain structures and improving discriminability. It achieves state-of-the-art (SOTA) performance in cross-resolution SAR target detection.

cs.CV↗

arXiv2025

Self-consistent clustering analysis for homogenisation of heterogeneous plates

This work introduces a reduced-order model for plate structures with periodic micro-structures by coupling self-consistent clustering analysis (SCA) with the Lippmann-Schwinger equation, enabling rapid multiscale homogenisation of heterogeneous plates. A plate-specific SCA scheme is derived for the first time and features two key elements: (i) an offline-online strategy that combines Green's functions with k-means data compression, and (ii) an online self-consistent update that exploits the weak sensitivity of the reference medium. The framework handles both linear and nonlinear problems in classical plate theory and first-order shear deformation theory, and its performance is verified on linear isotropic perforated plates and woven composites, as well as on non-linear elasto-plastic perforated plates and woven composites with damage. Across all cases the proposed model matches the accuracy of FFT-based direct numerical simulation while reducing computational cost by over an order of magnitude.

physics.comp-ph↗

arXiv2025

Finite-PINN: A Physics-Informed Neural Network with Finite Geometric Encoding for Solid Mechanics

PINN models have demonstrated capabilities in addressing fluid PDE problems, and their potential in solid mechanics is beginning to emerge. This study identifies two key challenges when using PINN to solve general solid mechanics problems. These challenges become evident when comparing the limitations of PINN with the well-established numerical methods commonly used in solid mechanics, such as the finite element method (FEM). Specifically: a) PINN models generate solutions over an infinite domain, which conflicts with the finite boundaries typical of most solid structures; and b) the solution space utilised by PINN is Euclidean, which is inadequate for addressing the complex geometries often present in solid structures. This work presents a PINN architecture for general solid mechanics problems, referred to as the Finite-PINN model. The model is designed to effectively tackle two key challenges, while retaining as much of the original PINN framework as possible. To this end, the Finite-PINN incorporates finite geometric encoding into the neural network inputs, thereby transforming the solution space from a conventional Euclidean space into a hybrid Euclidean-topological space. The model is comprehensively trained using both strong-form and weak-form loss formulations, enabling its application to a wide range of forward and inverse problems in solid mechanics. For forward problems, the Finite-PINN model efficiently approximates solutions to solid mechanics problems when the geometric information of a given structure has been preprocessed. For inverse problems, it effectively reconstructs full-field solutions from very sparse observations by embedding both physical laws and geometric information within its architecture.

cs.CE↗

arXiv2025

3D-HGS: 3D Half-Gaussian Splatting

Photo-realistic image rendering from 3D scene reconstruction has advanced significantly with neural rendering techniques. Among these, 3D Gaussian Splatting (3D-GS) outperforms Neural Radiance Fields (NeRFs) in quality and speed but struggles with shape and color discontinuities. We propose 3D Half-Gaussian (3D-HGS) kernels as a plug-and-play solution to address these limitations. Our experiments show that 3D-HGS enhances existing 3D-GS methods, achieving state-of-the-art rendering quality without compromising speed.

cs.CV↗

arXiv2025

Physics-informed Neural Network Predictive Control for Quadruped Locomotion

This study introduces a unified control framework that addresses the challenge of precise quadruped locomotion with unknown payloads, named as online payload identification-based physics-informed neural network predictive control (OPI-PINNPC). By integrating online payload identification with physics-informed neural networks (PINNs), our approach embeds identified mass parameters directly into the neural network's loss function, ensuring physical consistency while adapting to changing load conditions. The physics-constrained neural representation serves as an efficient surrogate model within our nonlinear model predictive controller, enabling real-time optimization despite the complex dynamics of legged locomotion. Experimental validation on our quadruped robot platform demonstrates 35% improvement in position and orientation tracking accuracy across diverse payload conditions (25-100 kg), with substantially faster convergence compared to previous adaptive control methods. Our framework provides a adaptive solution for maintaining locomotion performance under variable payload conditions without sacrificing computational efficiency.

cs.RO↗

arXiv2025

RAGFormer: Learning Semantic Attributes and Topological Structure for Fraud Detection

Fraud detection remains a challenging task due to the complex and deceptive nature of fraudulent activities. Current approaches primarily concentrate on learning only one perspective of the graph: either the topological structure of the graph or the attributes of individual nodes. However, we conduct empirical studies to reveal that these two types of features, while nearly orthogonal, are each independently effective. As a result, previous methods can not fully capture the comprehensive characteristics of the fraud graph. To address this dilemma, we present a novel framework called Relation-Aware GNN with transFormer~(RAGFormer) which simultaneously embeds both semantic and topological features into a target node. The simple yet effective network consists of a semantic encoder, a topology encoder, and an attention fusion module. The semantic encoder utilizes Transformer to learn semantic features and node interactions across different relations. We introduce Relation-Aware GNN as the topology encoder to learn topological features and node interactions within each relation. These two complementary features are interleaved through an attention fusion module to support prediction by both orthogonal features. Extensive experiments on two popular public datasets demonstrate that RAGFormer achieves state-of-the-art performance. The significant improvement of RAGFormer in an industrial credit card fraud detection dataset further validates the applicability of our method in real-world business scenarios.

cs.LG↗

arXiv2024

LoRKD: Low-Rank Knowledge Decomposition for Medical Foundation Models

The widespread adoption of large-scale pre-training techniques has significantly advanced the development of medical foundation models, enabling them to serve as versatile tools across a broad range of medical tasks. However, despite their strong generalization capabilities, medical foundation models pre-trained on large-scale datasets tend to suffer from domain gaps between heterogeneous data, leading to suboptimal performance on specific tasks compared to specialist models, as evidenced by previous studies. In this paper, we explore a new perspective called "Knowledge Decomposition" to improve the performance on specific medical tasks, which deconstructs the foundation model into multiple lightweight expert models, each dedicated to a particular anatomical region, with the aim of enhancing specialization and simultaneously reducing resource consumption. To accomplish the above objective, we propose a novel framework named Low-Rank Knowledge Decomposition (LoRKD), which explicitly separates gradients from different tasks by incorporating low-rank expert modules and efficient knowledge separation convolution. The low-rank expert modules resolve gradient conflicts between heterogeneous data from different anatomical regions, providing strong specialization at lower costs. The efficient knowledge separation convolution significantly improves algorithm efficiency by achieving knowledge separation within a single forward propagation. Extensive experimental results on segmentation and classification tasks demonstrate that our decomposed models not only achieve state-of-the-art performance but also exhibit superior transferability on downstream tasks, even surpassing the original foundation models in task-specific evaluations. The code is available at here.

cs.CV↗

arXiv2024

A MEMS-based terahertz broadband beam steering technique

A multi-level tunable reflection array wide-angle beam scanning method is proposed to address the limited bandwidth and small scanning angle issues of current terahertz beam scanning technology. In this method, a focusing lens and its array are used to achieve terahertz wave spatial beam control, and MEMS mirrors and their arrays are used to achieve wide-angle beam scanning. The 1~3 order terahertz MEMS beam scanning system designed based on this method can extend the mechanical scanning angle of MEMS mirrors by 2~6 times, when tested and verified using an electromagnetic MEMS mirror with a 7mm optical aperture and a scanning angle of 15° and a D-band terahertz signal source. The experiment shows that the operating bandwidth of the first-order terahertz MEMS beam scanning system is better than 40GHz, the continuous beam scanning angle is about 30°, the continuous beam scanning cycle response time is about 1.1ms, and the antenna gain is better than 15dBi at 160GHz. This method has been validated for its large bandwidth and scalable scanning angle, and has potential application prospects in terahertz dynamic communication, detection radar, scanning imaging, and other fields.

physics.optics↗

arXiv2024

Reprogramming Distillation for Medical Foundation Models

Medical foundation models pre-trained on large-scale datasets have demonstrated powerful versatile capabilities for various tasks. However, due to the gap between pre-training tasks (or modalities) and downstream tasks (or modalities), the real-world computation and speed constraints, it might not be straightforward to apply medical foundation models in the downstream scenarios. Previous methods, such as parameter efficient fine-tuning (PEFT) methods and knowledge distillation (KD) methods, are unable to simultaneously address the task (or modality) inconsistency and achieve personalized lightweight deployment under diverse real-world demands. To address the above issues, we propose a novel framework called Reprogramming Distillation (RD). On one hand, RD reprograms the original feature space of the foundation model so that it is more relevant to downstream scenarios, aligning tasks and modalities. On the other hand, through a co-training mechanism and a shared classifier, connections are established between the reprogrammed knowledge and the knowledge of student models, ensuring that the reprogrammed feature space can be smoothly mimic by the student model of different structures. Further, to reduce the randomness under different training conditions, we design a Centered Kernel Alignment (CKA) distillation to promote robust knowledge transfer. Empirically, we show that on extensive datasets, RD consistently achieve superior performance compared with previous PEFT and KD methods.

cs.CV↗

arXiv2024

Exploring Training on Heterogeneous Data with Mixture of Low-rank Adapters

Training a unified model to take multiple targets into account is a trend towards artificial general intelligence. However, how to efficiently mitigate the training conflicts among heterogeneous data collected from different domains or tasks remains under-explored. In this study, we explore to leverage Mixture of Low-rank Adapters (MoLA) to mitigate conflicts in heterogeneous data training, which requires to jointly train the multiple low-rank adapters and their shared backbone. Specifically, we introduce two variants of MoLA, namely, MoLA-Grad and MoLA-Router, to respectively handle the target-aware and target-agnostic scenarios during inference. The former uses task identifiers to assign personalized low-rank adapters to each task, disentangling task-specific knowledge towards their adapters, thereby mitigating heterogeneity conflicts. The latter uses a novel Task-wise Decorrelation (TwD) loss to intervene the router to learn oriented weight combinations of adapters to homogeneous tasks, achieving similar effects. We conduct comprehensive experiments to verify the superiority of MoLA over previous state-of-the-art methods and present in-depth analysis on its working mechanism. Source code is available at: https://github.com/MediaBrain-SJTU/MoLA

cs.CV↗

arXiv2024

Low-Rank Knowledge Decomposition for Medical Foundation Models

The popularity of large-scale pre-training has promoted the development of medical foundation models. However, some studies have shown that although foundation models exhibit strong general feature extraction capabilities, their performance on specific tasks is still inferior to task-specific methods. In this paper, we explore a new perspective called ``Knowledge Decomposition'' to improve the performance on specific medical tasks, which deconstruct the foundation model into multiple lightweight expert models, each dedicated to a particular task, with the goal of improving specialization while concurrently mitigating resource expenditure. To accomplish the above objective, we design a novel framework named Low-Rank Knowledge Decomposition (LoRKD), which explicitly separates graidents by incorporating low-rank expert modules and the efficient knowledge separation convolution. Extensive experimental results demonstrate that the decomposed models perform well in terms of performance and transferability, even surpassing the original foundation models.

cs.CV↗

arXiv2023

GPINN: Physics-informed Neural Network with Graph Embedding

This work proposes a Physics-informed Neural Network framework with Graph Embedding (GPINN) to perform PINN in graph, i.e. topological space instead of traditional Euclidean space, for improved problem-solving efficiency. The method integrates topological data into the neural network's computations, which significantly boosts the performance of the Physics-Informed Neural Network (PINN). The graph embedding technique infuses extra dimensions into the input space to encapsulate the spatial characteristics of a graph while preserving the properties of the original space. The selection of these extra dimensions is guided by the Fiedler vector, offering an optimised pathologic notation of the graph. Two case studies are conducted, which demonstrate significant improvement in the performance of GPINN in comparison to traditional PINN, particularly in its superior ability to capture physical features of the solution.

cs.LG↗

arXiv2022

The propagation of air fingers into an elastic branching network

We study experimentally the propagation of an air finger through the Y-bifurcation of an elastic, liquid-filled Hele-Shaw channel, as a benchtop model of airway reopening. With channel compliance provided by an elastic upper boundary, we can impose collapsed channel configurations into which we inject air with constant volume-flux. We typically observe steady finger propagation in the main channel, which is lost ahead of the Y-bifurcation but subsequently recovered in the daughter channels. At low levels of initial collapse, steady finger shapes and bubble pressure in the daughter channels map onto those in the main channel, despite small differences in initial collapse in different parts of the Y-channel. However, at higher levels of initial collapse where the elastic sheet almost touches the bottom boundary of the channel, experimentally indistinguishable fingers in the main channel can lead to multiple states of reopening of the daughter channels. The downstream distance at which steady propagation is recovered in the daughter channels also varies considerably with injection flow rate and initial collapse because of a transition in the mechanics regulating finger propagation. We find that the characteristic time and length-scales of this recovery are largest in the regime where viscous and surface tension forces dominate at low flow rate and/or low initial collapse, and that they decrease towards a constant plateau reached in the limit where elastic and surface tension forces balance at high flow rate and/or high initial collapse. Our findings suggest that practical networks are unlikely to comprise long enough channels for steady state propagation to remain established.

physics.flu-dyn↗

arXiv2021

Self-Supervised Decomposition, Disentanglement and Prediction of Video Sequences while Interpreting Dynamics: A Koopman Perspective

Human interpretation of the world encompasses the use of symbols to categorize sensory inputs and compose them in a hierarchical manner. One of the long-term objectives of Computer Vision and Artificial Intelligence is to endow machines with the capacity of structuring and interpreting the world as we do. Towards this goal, recent methods have successfully been able to decompose and disentangle video sequences into their composing objects and dynamics, in a self-supervised fashion. However, there has been a scarce effort in giving interpretation to the dynamics of the scene. We propose a method to decompose a video into moving objects and their attributes, and model each object's dynamics with linear system identification tools, by means of a Koopman embedding. This allows interpretation, manipulation and extrapolation of the dynamics of the different objects by employing the Koopman operator K. We test our method in various synthetic datasets and successfully forecast challenging trajectories while interpreting them.

cs.CV↗

arXiv2018

Probing Neutrino Dirac Mass in Left-Right Symmetric Models at the LHC and Next Generation Colliders

We assess the sensitivity of the LHC, its high energy upgrade, and a prospective 100 TeV hadronic collider to the Dirac Yukawa coupling of the heavy neutrinos in left-right symmetric models (LRSMs). We focus specifically on the trilepton final state in regions of parameter space yielding prompt decays of the right-handed gauge bosons ($W_R$) and neutrinos ($N_R$). In the minimal LRSM, the Dirac Yukawa couplings are completely fixed in terms of the mass matrices for the heavy and light neutrinos. In this case, the trilepton signal provides a direct probe of the Dirac mass term for a fixed $W_R$ and $N_R$ mass. We find that while it is possible to discover the $W_R$ at the LHC, probing the Dirac Yukawa couplings will require a 100 TeV $pp$ collider. We also show that the observation of the trilepton signal at the LHC would indicate the presence of a non-minimal LRSM scenario.

hep-ph↗

arXiv2018

High-speed PAM4-based Optical SDM Interconnects with Directly Modulated Long-wavelength VCSEL

This paper reports the demonstration of high-speed PAM-4 transmission using a 1.5-μm single-mode vertical cavity surface emitting laser (SM-VCSEL) over multicore fiber with 7 cores over different distances. We have successfully generated up to 70 Gbaud 4-level pulse amplitude modulation (PAM-4) signals with a VCSEL in optical back-to-back, and transmitted 50 Gbaud PAM-4 signals over both 1-km dispersion-uncompensated and 10-km dispersion-compensated in each core, enabling a total data throughput of 700 Gbps over the 7-core fiber. Moreover, 56 Gbaud PAM-4 over 1-km has also been shown, whereby unfortunately not all cores provide the required 3.8 $\times$ 10 $^{-3}$ bit error rate (BER) for the 7% overhead-hard decision forward error correction (7% OH HDFEC). The limited bandwidth of the VCSEL and the adverse chromatic dispersion of the fiber are suppressed with pre-equalization based on accurate end-to-end channel characterizations. With a digital post-equalization, BER performance below the 7% OH-HDFEC limit is achieved over all cores. The demonstrated results show a great potential to realize high-capacity and compact short-reach optical interconnects for data centers.

eess.SP↗