Searcharxiv⌕ Search

arXiv subjects

Wenchao Zhang

Publications and source records attributed to Wenchao Zhang.

At least 19 recordsLinked to original sources
arXiv2026

A Multimodal Pre-trained Network for Integrated EEG-Video Seizure Detection

Reliable seizure detection in mouse models is essential for preclinical epilepsy research, yet manual review of synchronized video-EEG recordings is labor-intensive and single-modality systems fail for complementary reasons: video-based methods are easily confounded by benign behaviors, whereas EEG-based methods are vulnerable to ictal motion artifacts. We present EEGVFusion, a multimodal framework that combines self-supervised EEG representation learning, spatio-temporal video encoding, optimal-transport alignment, and bidirectional cross-attention to integrate neural and behavioral evidence. We also curate an expert-annotated dataset of synchronized EEG and video recordings comprising 93 sessions from 15 mice for training and evaluation. In the random-session split, EEGVFusion achieved a Balanced Accuracy of 0.9957 with perfect event sensitivity and an Event FAR of 0.6250 FP/h, indicating strong seizure detection performance with a low false-alarm burden. In a single held-out-subject evaluation with Subject 110 reserved for testing, EEGVFusion achieved a Balanced Accuracy of 0.9718 and reduced Event FAR from 2.7250 FP/h for the EEG-only counterpart to 0.4833 FP/h while preserving perfect event sensitivity. Targeted ablations further showed that EEG pre-training and OT alignment help reduce false alarms while preserving event sensitivity.

cs.CV↗
arXiv2026

Deciphering the universal scaling of particle transverse momentum spectra in heavy-ion collisions

We systematically investigate the scaling properties of the transverse momentum spectra for pions, kaons, and protons in Au+Au collisions at $\sqrt{s_{NN}}$ = 7.7, 11.5, 14.5, 19.6, 27, 39, 62.4, and 200 GeV, as well as in U+U collisions at $\sqrt{s_{NN}}$ = 193 GeV, across different centrality classes, using experimental data from the collaborations at the Relativistic Heavy Ion Collider (RHIC). Universal scaling emerges when the particle transverse momentum spectra are scaled by global physical quantities, i.e., the average total particle multiplicity and mean transverse momentum, confirming recent scaling findings from the data at the Large Hadron Collider (LHC) by the ExTrEMe collaboration. The scaling behavior breaks down in the high $p_{T}$ region and in peripheral collisions. We provide a natural explanation for these observations by invoking the Cooper-Frye formula, which is used for hadronization in hydrodynamics. Furthermore, we demonstrate the equivalence between the scaling found by the ExTrEMe collaboration and the Hwa-Yang scaling which was proposed two decades ago.

hep-ph↗
arXiv2026

Efficient Deep Demosaicing with Spatially Downsampled Isotropic Networks

In digital imaging, image demosaicing is a crucial first step which recovers the RGB information from a color filter array (CFA). Oftentimes, deep learning is utilized to perform image demosaicing. Given that most modern digital imaging applications occur on mobile platforms, applying deep learning to demosaicing requires lightweight and efficient networks. Isotropic networks, also known as residual-in-residual networks, have been often employed for image demosaicing and joint-demosaicing-and-denoising (JDD). Most demosaicing isotropic networks avoid spatial downsampling entirely, and thus are often prohibitively expensive computationally for mobile applications. Contrary to previous isotropic network designs, this paper claims that spatial downsampling to a signficant degree can improve the efficiency and performance of isotropic networks. To validate this claim, we design simple fully convolutional networks with and without downsampling using a mathematical architecture design technique adapted from DeepMAD, and find that downsampling improves empirical performance. Additionally, empirical testing of the downsampled variant, JD3Net, of our fully convolutional networks reveals strong empirical performance on a variety of image demosaicing and JDD tasks.

cs.CV↗
arXiv2025

Revisiting Privacy-Utility Trade-off for DP Training with Pre-existing Knowledge

Differential privacy (DP) provides a provable framework for protecting individuals by customizing a random mechanism over a privacy-sensitive dataset. Deep learning models have demonstrated privacy risks in model exposure as an established learning model unintentionally records membership-level privacy leakage. Differentially private stochastic gradient descent (DP-SGD) has been proposed to safeguard training individuals by adding random Gaussian noise to gradient updates in the backpropagation. Researchers identify that DP-SGD causes utility loss since the injected homogeneous noise can alter the gradient updates calculated at each iteration. Namely, all elements in the gradient are contaminated regardless of their importance in updating model parameters. In this work, we argue that the utility can be optimized by involving the heterogeneity of the the injected noise. Consequently, we propose a generic differential privacy framework with heterogeneous noise (DP-Hero) by defining a heterogeneous random mechanism to abstract its property. The insight of DP-Hero is to leverage the knowledge encoded in the previously trained model to guide the subsequent allocation of noise heterogeneity, thereby leveraging the statistical perturbation and achieving enhanced utility. Atop DP-Hero, we instantiate a heterogeneous version of DP-SGD, and further extend it to federated training. We conduct comprehensive experiments to verify and explain the effectiveness of the proposed DP-Hero, showing improved training accuracy compared with state-of-the-art works. Broadly, we shed light on improving the privacy-utility space by learning the noise guidance from the pre-existing leaked knowledge encoded in the previously trained model, showing a different perspective of understanding the utility-improved DP training.

cs.CR↗
arXiv2025

Vectorial Symmetry Decoding with Single-Particle Precision via Room-Temperature Lanthanide Luminescence Polarimetry

Determining the local symmetry of luminescent centers in crystals is critical for understanding and controlling their optical transitions, yet current methods are limited by stringent experimental requirements and ambiguous symmetry assignments. Here, we develop a robust computational electromagnetics framework that directly connect the local symmetry and chirality of rare-earth-doped single crystals to the polarization states of their emitted light. This framework is experimentally validated through the precise determination of point and space group symmetries using high-resolution, polarization-resolved micro-photoluminescence (μ-PL) spectra. Unlike conventional approaches that usually rely on analyzing multiple transitions at cryogenic temperatures, our technique operates at room temperature, requires only a single optical transition, and enables accurate orientation of symmetry axes. This enables deterministic polarization control of nano-emitters by tailoring symmetry groups and selecting appropriate transition dipoles, eliminating the need for bulky or complex photonic structures. Additionally, we demonstrate the function of bio-sensing, via determining single particle orientations in complex cellular environments using minimal polarization measurements. These results pave the way for advances in energy transfer systems, ultra-bright rare-earth nanocrystals, nanophotonic materials, and real-time single-particle tracking in biological contexts.

physics.optics↗
arXiv2025

Enhancing Document-Level Question Answering via Multi-Hop Retrieval-Augmented Generation with LLaMA 3

This paper presents a novel Retrieval-Augmented Generation (RAG) framework tailored for complex question answering tasks, addressing challenges in multi-hop reasoning and contextual understanding across lengthy documents. Built upon LLaMA 3, the framework integrates a dense retrieval module with advanced context fusion and multi-hop reasoning mechanisms, enabling more accurate and coherent response generation. A joint optimization strategy combining retrieval likelihood and generation cross-entropy improves the model's robustness and adaptability. Experimental results show that the proposed system outperforms existing retrieval-augmented and generative baselines, confirming its effectiveness in delivering precise, contextually grounded answers.

cs.CL↗
arXiv2025

Align Beyond Prompts: Evaluating World Knowledge Alignment in Text-to-Image Generation

Recent text-to-image (T2I) generation models have advanced significantly, enabling the creation of high-fidelity images from textual prompts. However, existing evaluation benchmarks primarily focus on the explicit alignment between generated images and prompts, neglecting the alignment with real-world knowledge beyond prompts. To address this gap, we introduce Align Beyond Prompts (ABP), a comprehensive benchmark designed to measure the alignment of generated images with real-world knowledge that extends beyond the explicit user prompts. ABP comprises over 2,000 meticulously crafted prompts, covering real-world knowledge across six distinct scenarios. We further introduce ABPScore, a metric that utilizes existing Multimodal Large Language Models (MLLMs) to assess the alignment between generated images and world knowledge beyond prompts, which demonstrates strong correlations with human judgments. Through a comprehensive evaluation of 8 popular T2I models using ABP, we find that even state-of-the-art models, such as GPT-4o, face limitations in integrating simple real-world knowledge into generated images. To mitigate this issue, we introduce a training-free strategy within ABP, named Inference-Time Knowledge Injection (ITKI). By applying this strategy to optimize 200 challenging samples, we achieved an improvement of approximately 43% in ABPScore. The dataset and code are available in https://github.com/smile365317/ABP.

cs.CV↗
arXiv2025

Centralizers in Free Associative Algebras and Generic Matrices

This paper is concerned with the completion of the proof of the Bergman centralizer theorem by using generic matrices based on our previous quantization proof \cite{KBRZh}. Additionally, we establish that the algebra of generic matrices with characteristic coefficients is integrally closed.

math.RA↗
arXiv2025

Torus actions on free associative algebras, lifting and Białynicki-Birula type theorems

We examine the problem of the linearity of an algebraic torus action in the associative setting. We prove the free algebra analog of a classical theorem of BialynickiBirula, which establishes linearity of maximal torus action. Additionally, we formulate and prove linearity theorems for specific classes of regular actions, and provide a framework for constructing non-linearizable actions, analogous to the work of Asanuma. This framework has applications in the study of the Associative Cancellation Conjecture. Furthermore, we show the existence of two non-isomorphic algebras, whose free products with a polynomial ring are isomorphic.

math.AG↗
arXiv2021

Joint Object Contour Points and Semantics for Instance Segmentation

The attributes of object contours has great significance for instance segmentation task. However, most of the current popular deep neural networks do not pay much attention to the object edge information. Inspired by the human annotation process when making instance segmentation datasets, in this paper, we propose Mask Point R-CNN aiming at promoting the neural network's attention to the object boundary. Specifically, we innovatively extend the original human keypoint detection task to the contour point detection of any object. Based on this analogy, we present an contour point detection auxiliary task to Mask R-CNN, which can boost the gradient flow between different tasks by effectively using feature fusion strategies and multi-task joint training. As a consequence, the model will be more sensitive to the edges of the object and can capture more geometric features. Quantitatively, the experimental results show that our approach outperforms vanilla Mask R-CNN by 3.8\% on Cityscapes dataset and 0.8\% on COCO dataset.

cs.CV↗
arXiv2021

A More Compact Object Detector Head Network with Feature Enhancement and Relational Reasoning

Modeling implicit feature interaction patterns is of significant importance to object detection tasks. However, in the two-stage detectors, due to the excessive use of hand-crafted components, it is very difficult to reason about the implicit relationship of the instance features. To tackle this problem, we analyze three different levels of feature interaction relationships, namely, the dependency relationship between the cropped local features and global features, the feature autocorrelation within the instance, and the cross-correlation relationship between the instances. To this end, we propose a more compact object detector head network (CODH), which can not only preserve global context information and condense the information density, but also allows instance-wise feature enhancement and relational reasoning in a larger matrix space. Without bells and whistles, our method can effectively improve the detection performance while significantly reducing the parameters of the model, e.g., with our method, the parameters of the head network is 0.6 times smaller than the state-of-the-art Cascade R-CNN, yet the performance boost is 1.3% on COCO test-dev. Without losing generality, we can also build a more lighter head network for other multi-stage detectors by assembling our method.

cs.CV↗
arXiv2020

Global Context Aware RCNN for Object Detection

RoIPool/RoIAlign is an indispensable process for the typical two-stage object detection algorithm, it is used to rescale the object proposal cropped from the feature pyramid to generate a fixed size feature map. However, these cropped feature maps of local receptive fields will heavily lose global context information. To tackle this problem, we propose a novel end-to-end trainable framework, called Global Context Aware (GCA) RCNN, aiming at assisting the neural network in strengthening the spatial correlation between the background and the foreground by fusing global context information. The core component of our GCA framework is a context aware mechanism, in which both global feature pyramid and attention strategies are used for feature extraction and feature refinement, respectively. Specifically, we leverage the dense connection to improve the information flow of the global context at different stages in the top-down process of FPN, and further use the attention mechanism to refine the global context at each level in the feature pyramid. In the end, we also present a lightweight version of our method, which only slightly increases model complexity and computational burden. Experimental results on COCO benchmark dataset demonstrate the significant advantages of our approach.

cs.CV↗
arXiv2020

Nonstandard analysis, deformation quantization and some logical aspects of (non)commutative algebraic geometry

This paper surveys results related to well-known works of B. Plotkin and V. Remeslennikov on the edge of algebra, logic and geometry. We start from a brief review of the paper and motivations. The first sections deal with model theory. In Section 2.1 we describe the geometric equivalence, the elementary equivalence, and the isotypicity of algebras. We look at these notions from the positions of universal algebraic geometry and make emphasis on the cases of the first order rigidity. In this setting Plotkin's problem on the structure of automorphisms of (auto)endomorphisms of free objects, and auto-equivalence of categories is pretty natural and important. Section 2.2 is dedicated to particular cases of Plotkin's problem. Section 2.3 is devoted to Plotkin's problem for automorphisms of the group of polynomial symplectomorphisms. This setting has applications to mathematical physics through the use of model theory (non-standard analysis) in the studying of homomorphisms between groups of symplectomorphisms and automorphisms of the Weyl algebra. The last two sections deal with algorithmic problems for noncommutative and commutative algebraic geometry. Section 3.1 is devoted to the Gröbner basis in non-commutative situation. Despite the existence of an algorithm for checking equalities, the zero divisors and nilpotency problems are algorithmically unsolvable. Section 3.2 is connected with the problem of embedding of algebraic varieties; a sketch of the proof of its algorithmic undecidability over a field of characteristic zero is given.

math.RA↗
arXiv2019

Nonvolatile Multilevel States in Multiferroic Tunnel Junctions

Manipulation of tunneling spin-polarized electrons via a ferroelectric interlayer sandwiched between two ferromagnetic electrodes, dubbed Multiferroic Tunnel Junctions (MFTJs), can be achieved not only by the magnetic alignments of two ferromagnets but also by the electric polarization of the ferroelectric interlayer, providing great opportunities for next-generation multi-state memory devices. Here we show that a La0.67Sr0.33MnO3 (LSMO)/PbZr0.2Ti0.8O3(PZT)/Co structured MFTJ device can exhibit multilevel resistance states in the presence of gradually reversed ferroelectric domains via tunneling electro-resistance and tunneling magnetoresistance, respectively. The nonvolatile ferroelectric control in the MFTJ can be attributed to separate contributions arising from two independent ferroelectric channels in the PZT interlayer with opposite polarization. Our study shows the dominant role of "mixed" ferroelectric states on achieving accumulative electrical modulation of multilevel resistance states in MFTJs, paving the way for multifunctional device applications.

cond-mat.mtrl-sci↗
arXiv2019

Noncommutative Bialynicki-Birula Theorem

In this short note we prove that every maximal torus action on the free algebra is conjugate to a linear action. This statement is the free algebra analogue of a classical theorem of A. Białynicki-Birula.

math.AG↗
arXiv2018

FTIR and GCMS analysis of epoxy resin decomposition products feeding the flame during UL 94 standard flammability test. Application to the understanding of the blowing-out effect in epoxy/polyhedral silsesquioxane formulations

A novel method was developed for the sampling of volatiles produces by polymer decomposition during UL94 standard flammability tests, allowing to collect, separate and analyze the precise composition of the fuel mixture feeding the flame in the real flammability tests. The system was validated on epoxy resin/polyhedral oligomeric silsesquioxanes and found extremely informative for the understanding of the flame retardancy mechanisms of POSS, previously referred to as the blowing-out effect. Collected products were analyzed by infrared spectroscopy and gas chromatography/mass spectroscopy, to identify the gaseous and liquid decomposition products, allowing to depict a comprehensive decomposition pathway for the epoxy resin. Lightweight volatiles, gaseous at room temperature, showed limited differences as a function of DOPO-POSS, whereas mixtures of liquid products evidenced for dramatic changes in the relative concentration as a function of DOPO-POSS. In pristine epoxy resin, the most abundant products were recognized as benzene, phenol, naphthalene and toluene, along with several tens of other aromatic products observed in lower amounts. The presence of DOPO-POSS at low concentration radically changes the composition of the aromatic volatiles mixture, as no significant amount of benzene is produced, while phenol becomes the main product, accounting for about half of the total, isopropyl phenol and bisphenol A, along with several tens of other products in lower concentrations. Such modification of the fuel mixture feeding the flame provided an explanation for the lower flammability of this formulation, as well as for the phenomenology of the blowing out effect.

physics.chem-ph↗
arXiv2018

Universal scaling of strange particle $p_{\rm T}$ spectra in pp collisions

As a complementary study to that performed on the transverse momentum ($p_{\rm T}$) spectra of charged pions, kaons and protons in proton-proton (pp) collisions at LHC energies 0.9, 2.76 and 7 TeV, we present a scaling behaviour in the $p_{\rm T}$ spectra of strange particles ($K_{S}^{0}$, $\rm Λ$, $\rm Ξ$ and $ϕ$) at these three energies. This scaling behaviour is exhibited when the spectra are expressed in a suitable scaling variable $z=p_{\rm T}/K$, where the scaling parameter $K$ is determined by the quality factor method and increases with the center of mass energy ($\sqrt{s}$). The rates at which $K$ increases with $\mathrm{ln}\sqrt{s}$ for these strange particles are found to be identical within errors. In the framework of the colour string percolation model, we argue that these strange particles are produced through the decay of clusters that are formed by the colour strings overlapping. We observe that the strange mesons and baryons are produced from clusters with different size distributions, while the strange mesons (baryons) $K_{S}^{0}$ and $ϕ$ ($\rm Λ$ and $\rm Ξ$) originate from clusters with the same size distributions. The cluster's size distributions for strange mesons are more dispersed than those for strange baryons. The scaling behaviour of the $p_{\rm T}$ spectra for these strange particles can be explained by the colour string percolation model in a quantitative way.

hep-ph↗