arXiv subjects

Wei Cao

Publications and source records attributed to Wei Cao.

At least 37 records · Page 2Linked to original sources

arXiv2026

Renewable estimation in linear expectile regression models with streaming data sets

Streaming data often exhibit heterogeneity due to heteroscedastic variances or inhomogeneous covariate effects. Online renewable quantile and expectile regression methods provide valuable tools for detecting such heteroscedasticity by combining current data with summary statistics from historical data. However, quantile regression can be computationally demanding because of the non-smooth check function. To address this, we propose a novel online renewable method based on expectile regression, which efficiently updates estimates using both current observations and historical summaries, thereby reducing storage requirements. By exploiting the smoothness of the expectile loss function, our approach achieves superior computational efficiency compared with existing online renewable methods for streaming data with heteroscedastic variances or inhomogeneous covariate effects. We establish the consistency and asymptotic normality of the proposed estimator under mild regularity conditions, demonstrating that it achieves the same statistical efficiency as oracle estimators based on full individual-level data. Numerical experiments and real-data applications demonstrate that our method performs comparably to the oracle estimator while maintaining high computational efficiency and minimal storage costs.

stat.ME↗

arXiv2026

Robust optimal reconciliation for hierarchical time series forecasting with M-estimation

Aggregation constraints, arising from geographical or sectoral division, frequently emerge in a large set of time series. Coherent forecasts of these constrained series are anticipated to conform to their hierarchical structure organized by the aggregation rules. To enhance its resilience against potential irregular series, we explore the robust reconciliation process for hierarchical time series (HTS) forecasting. We incorporate M-estimation to obtain the reconciled forecasts by minimizing a robust loss function of transforming a group of base forecasts subject to the aggregation constraints. The related minimization procedure is developed and implemented through a modified Newton-Raphson algorithm via local quadratic approximation. Extensive numerical experiments are carried out to evaluate the performance of the proposed method, and the results suggest its feasibility in handling numerous abnormal cases (for instance, series with non-normal errors). The proposed robust reconciliation also demonstrates excellent efficiency when no outliers exist in HTS. Finally, we showcase the practical application of the proposed method in a real-data study on Australian domestic tourism.

stat.AP↗

arXiv2025

Single-Axis Ptychographic Coherent Diffractive Imaging for Spectroscopic and Wavefront Retrieval

We present a novel axial ptychographic coherent diffractive imaging (AP-CDI) technique designed to overcome the critical throughput bottleneck of conventional methods. By replacing the 2D raster scan with a simple 1D axial scan, our approach reduces the number of required diffraction patterns by approximately an order of magnitude while maintaining high-fidelity reconstruction. We have experimentally validated this concept, successfully performing simultaneous spectroscopic imaging of a sample and quantitative wavefront characterization of the illumination, thereby accurately quantifying the chromatic aberration of a broadband field. This capability establishes AP-CDI as a highly efficient and versatile tool for real-time, multi-modal imaging, with immediate potential in ultrafast science, material characterization, and live-cell bio-imaging.

physics.optics↗

arXiv2025

Versatile and Efficient Medical Image Super-Resolution Via Frequency-Gated Mamba

Medical image super-resolution (SR) is essential for enhancing diagnostic accuracy while reducing acquisition cost and scanning time. However, modeling both long-range anatomical structures and fine-grained frequency details with low computational overhead remains challenging. We propose FGMamba, a novel frequency-aware gated state-space model that unifies global dependency modeling and fine-detail enhancement into a lightweight architecture. Our method introduces two key innovations: a Gated Attention-enhanced State-Space Module (GASM) that integrates efficient state-space modeling with dual-branch spatial and channel attention, and a Pyramid Frequency Fusion Module (PFFM) that captures high-frequency details across multiple resolutions via FFT-guided fusion. Extensive evaluations across five medical imaging modalities (Ultrasound, OCT, MRI, CT, and Endoscopic) demonstrate that FGMamba achieves superior PSNR/SSIM while maintaining a compact parameter footprint ($<$0.75M), outperforming CNN-based and Transformer-based SOTAs. Our results validate the effectiveness of frequency-aware state-space modeling for scalable and accurate medical image enhancement.

cs.CV↗

arXiv2025

Neural Networks for Censored Expectile Regression Based on Data Augmentation

Expectile regression neural networks (ERNNs) are powerful tools for capturing heterogeneity and complex nonlinear structures in data. However, most existing research has primarily focused on fully observed data, with limited attention paid to scenarios involving censored observations. In this paper, we propose a data augmentation based ERNNs algorithm, termed DAERNN, for modeling heterogeneous censored data. The proposed DAERNN is fully data driven, requires minimal assumptions, and offers substantial flexibility. Simulation studies and real data applications demonstrate that DAERNN outperforms existing censored ERNNs methods and achieves predictive performance comparable to models trained on fully observed data. Moreover, the algorithm provides a unified framework for handling various censoring mechanisms without requiring explicit parametric model specification, thereby enhancing its applicability to practical censored data analysis.

stat.ML↗

arXiv2025

Accelerating the Discovery of Materials with Expected Thermal Conductivity via a Synergistic Strategy of DFT and Interpretable Deep Learning

Lattice thermal conductivity (LTC) is a critical parameter for thermal transport properties, playing a pivotal role in advancing thermoelectric materials and thermal management technologies. Traditional computational methods, such as Density Functional Theory (DFT) and Molecular Dynamics (MD), are resource-intensive, limiting their applicability for high-throughput LTC prediction. While AI-driven approaches have made significant strides in material science, the trade-off between accuracy and interpretability remains a major bottleneck. In this study, we introduce an interpretable deep learning framework that enables rapid and accurate LTC prediction, effectively bridging the gap between interpretability and precision. Leveraging this framework, we identify and validate four promising thermal conductors/insulators using DFT and MD. Moreover, by combining sensitivity analysis with DFT calculations, we uncover novel insights into phonon thermal transport mechanisms, providing a deeper understanding of the underlying physics. This work not only accelerates the discovery of thermal materials but also sets a new benchmark for interpretable AI in material science.

cond-mat.mtrl-sci↗

arXiv2025

Deep Learning for Multivariate Time Series Imputation: A Survey

Missing values are ubiquitous in multivariate time series (MTS) data, posing significant challenges for accurate analysis and downstream applications. In recent years, deep learning-based methods have successfully handled missing data by leveraging complex temporal dependencies and learned data distributions. In this survey, we provide a comprehensive summary of deep learning approaches for multivariate time series imputation (MTSI) tasks. We propose a novel taxonomy that categorizes existing methods based on two key perspectives: imputation uncertainty and neural network architecture. Furthermore, we summarize existing MTSI toolkits with a particular emphasis on the PyPOTS Ecosystem, which provides an integrated and standardized foundation for MTSI research. Finally, we discuss key challenges and future research directions, which give insight for further MTSI research. This survey aims to serve as a valuable resource for researchers and practitioners in the field of time series analysis and missing data imputation tasks.A well-maintained MTSI paper and tool list are available at https://github.com/WenjieDu/Awesome_Imputation.

cs.LG↗

arXiv2025

Accelerating Multi-Objective Collaborative Optimization of Doped Thermoelectric Materials via Artificial Intelligence

The thermoelectric performance of materials exhibits complex nonlinear dependencies on both elemental types and their proportions, rendering traditional trial-and-error approaches inefficient and time-consuming for material discovery. In this work, we present a deep learning model capable of accurately predicting thermoelectric properties of doped materials directly from their chemical formulas, achieving state-of-the-art performance. To enhance interpretability, we further incorporate sensitivity analysis techniques to elucidate how physical descriptors affect the thermoelectric figure of merit (zT). Moreover, we establish a coupled framework that integrates a surrogate model with a multi-objective genetic algorithm to efficiently explore the vast compositional space for high-performance candidates. Experimental validation confirms the discovery of a novel thermoelectric material with superior $zT$ values in the medium-temperature regime.

cond-mat.mtrl-sci↗

arXiv2024

Interface and spectrum multiplexing ptychographic reflection microscopy

Reflective ptychography is a promising lensless imaging technique with a wide field of view, offering significant potential for applications in semiconductor manufacturing and detection. However, many semiconductor materials are coated with different layers during processing, which leads to the reflected diffraction light being a coherent superposition of multiple light beams. Traditional phase recovery methods often overlook the multi-layered nature of these materials, resulting in artificial errors and, in some cases, failures in image reconstruction. This limitation has hindered the broader application and adoption of reflection ptychography. Here, we propose and experimentally demonstrate an innovative interface and spectrum multiplexing ptychographic reflection microscopy. By employing multi-wavelength light as the illumination source, our approach enables the accurate extraction of the wavelength-dependent surface structure imaging and topography mapping of materials in a single experiment. This advancement offers a reliable technique for element-specific detection of semiconductor materials, utilizing tabletop extreme ultraviolet light sources in the future.

physics.optics↗

arXiv2024

3D Extended Target Sensing in ISAC: Cramér-Rao Bound Analysis and Beamforming Design

This paper investigates an integrated sensing and communication (ISAC) system where the sensing target is a three-dimensional (3D) extended target, for which multiple scatterers from the target surface can be resolved. We first introduce a second-order truncated Fourier series surface model for an arbitrarily-shaped 3D ET. Utilizing this model, we derive tractable Cramer-Rao bounds (CRBs) for estimating the ET kinematic parameters, including the center range, azimuth, elevation, and orientation. These CRBs depend explicitly on the transmit covariance matrix and ET shape. Then we formulate two transmit beamforming optimization problems for the base station (BS) to simultaneously support communication with multiple users and sensing of the 3D ET. The first minimizes the sensing CRB while ensuring a minimum signal-to-interference-plus-noise ratio (SINR) for each user, and it is solved using semidefinite relaxation. The second balances minimizing the CRB and maximizing communication rates through a weight factor, and is solved via successive convex approximation. To reduce the computational complexity, we further propose ISACBeam-GNN, a novel graph neural network-based beamforming method that employs a separate-then-integrate structure, learning communication and sensing (C&S) objectives independently before integrating them to balance C&S trade-offs. Simulation results show that the proposed beamforming designs that account for ET shapes significantly outperform existing baselines, offering better communication-sensing performance trade-offs as well as an improved beampattern for sensing. Results also demonstrate that ISACBeam-GNN is an efficient alternative to the optimization-based methods, with remarkable adaptability and scalability.

cs.IT↗

arXiv2024

Non-Hermitian quasicrystalline topological insulators

In recent years, the interplay between non-Hermiticity and band topology is expected to uncover numerous novel physical phenomena. However, the majority of research has focused on periodic crystalline structures, with comparatively fewer studies exploring quasicrystalline systems. In this paper, we delve into the influence of asymmetric hopping on the topological insulators, specifically focusing on quantum spin Hall insulators and higher-order topological insulators in an octagonal Ammann-Beenker quasicrystalline lattice. We demonstrate that asymmetric hopping can significantly alter the distribution of edge states, leading to a uniform distribution across all boundaries or localizing them at a single edge, depending on the symmetry adjustments. Furthermore, we explore the robustness of higher-order topological corner states under perturbations, showing that these states can maintain their distribution even in the presence of non-Hermiticity. Our findings not only expand the current understanding of topological states in quasicrystals under non-Hermitian conditions, but also provide valuable theoretical guidance for manipulating corner state distributions in non-Hermitian quasicrystalline higher-order topological insulators.

cond-mat.mes-hall↗

arXiv2024

Hybrid SIC Aided Hybrid NOMA: A New Approach For Improving Energy Efficiency

Hybrid non-orthogonal multiple access (NOMA), which organically combines pure NOMA and conventional OMA, has recently received significant attention to be a promising multiple access framework for future wireless communication networks. However, most of the literatures on hybrid NOMA only consider fixed order of successive interference cancellation (SIC), namely FSIC, for the NOMA transmission phase of hybrid NOMA, resulting in limited performance. Differently, this paper aims to reveal the potential of applying hybrid SIC (HSIC) to improve the energy efficiency of hybrid NOMA. Specifically, a HSIC aided hybrid NOMA scheme is proposed, which can be treated as a simple add-on to the legacy orthogonal multiple access (OMA) based network. The proposed scheme offers some users (termed ``opportunistic users'') to have more chances to transmit by transparently sharing legacy users' time slots. For a fair comparison, a power reducing coefficient $β$ is introduced to ensure that the energy consumption of the proposed scheme is less than conventional OMA. Given $β$, the probability for the event that the achievable rate of the proposed HSIC aided hybrid NOMA scheme cannot outperform its OMA counterpart is obtained in closed-form, by considering impact of user pairing. Furthermore, asymptotic analysis shows that the aforementioned probability can approach zero under some given conditions in the SNR regime, indicating that the energy efficiency of the proposed scheme is almost surely higher than that of OMA for these given conditions. Numerical results are presented to verify the analysis and also demonstrate the benefit of applying HSIC compared to FSIC.

cs.IT↗

arXiv2024

Spatio-spectral light modulator of XUV high harmonics

High-order harmonic generation (HHG), characterized by its highly nonlinear nature, often exhibits a complex spatio-temporal profile that poses challenges for practical applications. In this study, we unveil a method for manipulating the spatio-spectral distribution of HHG by guiding the recollision electron trajectory in the spatio-temporal domain using a control field. The resulting far-field high harmonic (HH) radiation inherits the intricate spatio-temporal characteristics of the control field, showcasing diverse features including spatial tilting, spectral shifting, and emission angle deflection. Using the relative delay between the control field and the driving pulse as the primary control parameter, we achieve precise tailoring of the high harmonics in the spatio-spectral domain. This controllability in HH benefits ultrafast metrology and imaging applications in the extreme ultraviolet (XUV) regime.

physics.optics↗

arXiv2024

What Matters to Enhance Traffic Rule Compliance of Imitation Learning for End-to-End Autonomous Driving

End-to-end autonomous driving, where the entire driving pipeline is replaced with a single neural network, has recently gained research attention because of its simpler structure and faster inference time. Despite this appealing approach largely reducing the complexity in the driving pipeline, it also leads to safety issues because the trained policy is not always compliant with the traffic rules. In this paper, we proposed P-CSG, a penalty-based imitation learning approach with contrastive-based cross semantics generation sensor fusion technologies to increase the overall performance of end-to-end autonomous driving. In this method, we introduce three penalties - red light, stop sign, and curvature speed penalty to make the agent more sensitive to traffic rules. The proposed cross semantics generation helps to align the shared information of different input modalities. We assessed our model's performance using the CARLA Leaderboard - Town 05 Long Benchmark and Longest6 Benchmark, achieving 8.5% and 2.0% driving score improvement compared to the baselines. Furthermore, we conducted robustness evaluations against adversarial attacks like FGSM and Dot attacks, revealing a substantial increase in robustness compared to other baseline models. More detailed information can be found at https://hk-zh.github.io/p-csg-plus.

cs.CV↗

arXiv2024

Polarization Saturation in Multi-layered Interfacial Ferroelectrics

Van der Waals (vdW) polytypes of broken inversion and mirror symmetries were recently shown to exhibit switchable electric polarization even at the ultimate two-layer thin limit. Their out-of-plane polarization was found to accumulate in a ladder-like fashion with each successive layer, offering 2D building blocks for the bottom-up construction of 3D ferroelectrics. Here, we demonstrate experimentally that beyond a critical stack thickness, the accumulated polarization in rhombohedral polytypes of molybdenum disulfide (r-MoS2) saturates. The underlying saturation mechanism, deciphered via DFT and self-consistent Poisson-Schrödinger calculations, point to a purely electronic redistribution involving: (1) polarization-induced bandgap closure that allows for cross-stack charge transfer and the emergence of free surface charge; (2) reduction of the polarization saturation value, as well as the critical thickness at which it is obtained, by the presence of free carriers. The resilience of polar layered structures to atomic surface reconstruction, which is essentially unavoidable in polar 3D crystals, potentially allows for the design of new devices with mobile surface charges. Our findings, which are of general nature, should be accounted for when designing switching and/or conductive devices based on ferroelectric layered materials.

cond-mat.mtrl-sci↗

arXiv2024

NuTime: Numerically Multi-Scaled Embedding for Large-Scale Time-Series Pretraining

Recent research on time-series self-supervised models shows great promise in learning semantic representations. However, it has been limited to small-scale datasets, e.g., thousands of temporal sequences. In this work, we make key technical contributions that are tailored to the numerical properties of time-series data and allow the model to scale to large datasets, e.g., millions of temporal sequences. We adopt the Transformer architecture by first partitioning the input into non-overlapping windows. Each window is then characterized by its normalized shape and two scalar values denoting the mean and standard deviation within each window. To embed scalar values that may possess arbitrary numerical amplitudes in a high-dimensional space, we propose a numerically multi-scaled embedding module enumerating all possible numerical scales for the scalars. The model undergoes pretraining with a simple contrastive objective on a large-scale dataset over a million sequences collected by merging existing public data. We study its transfer performance on a number of univariate and multivariate classification tasks, few shot learning, unsupervised clustering and anomaly detection benchmarks. Our method exhibits remarkable improvement against previous pretraining approaches and establishes the new state of the art, even compared with domain-specific non-learning-based methods. Code is available at: \url{https://github.com/chenguolin/NuTime}.

cs.LG↗

arXiv2024

Spectral multiplexing based on multi-distance lensless imaging

We have demonstrated the capability of spectral multiplexing in multi-distance diffractive imaging, enabling the reconstruction of samples with diverse spectral responses. While previous methods like ptychography utilize redundancy in radial diffraction data to achieve information multiplexing, they typically require capturing a substantial amount of diffraction data. In contrast, our approach effectively harnesses the redundancy information in axial diffraction data. This significantly reduces the amount of diffraction data required and relaxes the stringent requirements on optical path stability.

physics.optics↗

arXiv2024

Motion2VecSets: 4D Latent Vector Set Diffusion for Non-rigid Shape Reconstruction and Tracking

We introduce Motion2VecSets, a 4D diffusion model for dynamic surface reconstruction from point cloud sequences. While existing state-of-the-art methods have demonstrated success in reconstructing non-rigid objects using neural field representations, conventional feed-forward networks encounter challenges with ambiguous observations from noisy, partial, or sparse point clouds. To address these challenges, we introduce a diffusion model that explicitly learns the shape and motion distribution of non-rigid objects through an iterative denoising process of compressed latent representations. The diffusion-based priors enable more plausible and probabilistic reconstructions when handling ambiguous inputs. We parameterize 4D dynamics with latent sets instead of using global latent codes. This novel 4D representation allows us to learn local shape and deformation patterns, leading to more accurate non-linear motion capture and significantly improving generalizability to unseen motions and identities. For more temporally-coherent object tracking, we synchronously denoise deformation latent sets and exchange information across multiple frames. To avoid computational overhead, we designed an interleaved space and time attention block to alternately aggregate deformation latents along spatial and temporal domains. Extensive comparisons against state-of-the-art methods demonstrate the superiority of our Motion2VecSets in 4D reconstruction from various imperfect observations. More detailed information can be found at https://vveicao.github.io/projects/Motion2VecSets/.

cs.CV↗