Searcharxiv⌕ Search

arXiv subjects

Jie Du

Publications and source records attributed to Jie Du.

At least 19 recordsLinked to original sources
arXiv2026

Large Exchange Magnetostriction in a Kagome Antiferromagnet at Room Temperature

The pursuit of high-performance magnetostrictive materials is crucial for advancing technologies in sensing, actuation, and microelectromechanical systems. Although pronounced magnetostrictive effects have been observed in a few ferromagnets, a systematic exploration of magnetostriction across a broader range of antiferromagnets remains limited. Here, we report the observation of large magnetostriction in the kagome antiferromagnet YMn6Sn6. Under a magnetic field, the system undergoes a field-induced evolution from a helical magnetic ground state toward a collinear field-polarized state, accompanied by a large anisotropic lattice strain and a substantial volume magnetostriction exceeding 400 ppm at room temperature. Remarkably, the magnetostrictive response remains nearly fully reversible up to 9 T with nearly hysteresis-free behavior, effectively minimizing the energy dissipation commonly associated with domain-wall pinning. Combined experimental measurements and theoretical calculations reveal that the large, nearly hysteresis-free magnetostriction originates from the competition between intralayer ferromagnetic and interlayer exchange interactions. This exchange-driven magnetoelastic coupling further gives rise to a strongly direction-dependent lattice distortion pathway. This work establishes kagome helimagnets as a tunable platform for low-dissipation magnetoelastic functionalities and responsive magnetomechanical applications.

cond-mat.mtrl-sci↗
arXiv2026

Masked Generative-Contrastive Representation Learning for Cross-Dataset EEG-Based Emotion Recognition

Self-supervised learning (SSL) shows strong potential for cross-dataset transfer by improving feature representation and generalization. However, its application to EEG-based emotion recognition remains largely unexplored. Existing SSL methods struggle to capture the intricate spatiotemporal dependencies of EEG signals under varying channel configurations, extract fine-grained representations resilient to noise, and derive global features that generalize well across subjects. To address these challenges, we propose Masked Generative-Contrastive Representation Learning (MGCRL), a novel SSL framework specifically designed for EEG-based emotion recognition. Built upon a region-aware spatiotemporal encoder, MGCRL integrates generative and contrastive learning to achieve both fine-grained and global discriminative representations for cross-dataset generalization. MGCRL introduces three key designs: 1) a spatiotemporal encoder that incorporates region-based graph convolution to capture localized spatial and functional relationships, enhancing region-specific feature learning and mitigating the impact of varying EEG channel configurations across datasets; 2) a generative learning mechanism based on the joint embedding predictive architecture (JEPA) that utilizes masked features to capture noise robustness fine-grained representations, improving the model's capability to characterize subtle emotional states; and 3) a contrastive learning strategy that leverages masked and original features to learn temporally stable and cross-subject-invariant representations across the same stimuli, boosting emotion discrimination and cross-subject generalization. Under these designs, MGCRL exhibits remarkable ability to learn universal representation. Extensive experiments involving pretraining on the large FACED dataset and fine-tuning on multiple SEED-series datasets demonstrate the effectiveness of MGCRL.

cs.LG↗
arXiv2026

Giant Magnetostriction by Design: A First-Principles Screening of Co-based Heusler Alloys

The pursuit of high-performance, rare-earth-free magnetostrictive materials is crucial for advancing technologies in sensing, actuation, and microelectromechanical systems. Heusler alloys represent a promising, yet underexplored, class of materials for this purpose. In this work, we perform a systematic first-principles investigation of the magnetostrictive properties of 25 Co-based full Heusler alloys, Co$_2$YZ (Y = V, Cr, Mn, Fe, Co; Z = Al, Ga, Si, Ge, Sn). Our screening identifies 10 compounds with large predicted magnetostriction ($|\lambda_{001}| > 100$~ppm), highlighted by Co$_3$Si with a giant value of -966~ppm. Furthermore, we demonstrate two effective strategies for engineering magnetostriction: (i) tuning the Fermi level, which enhances the magnetostriction of Co$_3$Sn to -905~ppm via Sb doping, and (ii) amplifying the spin-orbit coupling, which boosts the magnetostriction of Co$_2$CrGa to a colossal -1008~ppm through Re substitution. Our analysis reveals a general predictive rule, uncovering a linear relationship between the magnetostriction and the choice of the Y-site transition metal. This work not only identifies novel candidates for magnetostrictive applications but also establishes clear, physically-grounded design principles to accelerate the discovery of new functional magnetic materials.

cond-mat.mtrl-sci↗
arXiv2026

From microscopic social force models to macroscopic continuum models for pedestrian flow

The pedestrian flow is one of the most complex systems, involving large populations of interacting agents. Models at microscopic and macroscopic scales offer different advantages for studying related problems. In general, microscopic models can describe interaction forces at the individual level. Macroscopic models, on the other hand, provide analytical insights into global interactions and long-term overall dynamics, along with efficient numerical simulations and predictions. However, the relationship between models at different scales has rarely been explored. In this study, based on the original microscopic social force model with a reactive optimal route choice strategy, we first derive kinetic equations at the mesoscopic level. By varying the interaction force in different scenarios, we then derive several continuum models at the macroscopic level. Finally, numerical examples are given to evaluate the behaviors of the social force model and our continuum models.

nlin.AO↗
arXiv2025

Reg-DPO: SFT-Regularized Direct Preference Optimization with GT-Pair for Improving Video Generation

Recent studies have identified Direct Preference Optimization (DPO) as an efficient and reward-free approach to improving video generation quality. However, existing methods largely follow image-domain paradigms and are mainly developed on small-scale models (approximately 2B parameters), limiting their ability to address the unique challenges of video tasks, such as costly data construction, unstable training, and heavy memory consumption. To overcome these limitations, we introduce a GT-Pair that automatically builds high-quality preference pairs by using real videos as positives and model-generated videos as negatives, eliminating the need for any external annotation. We further present Reg-DPO, which incorporates the SFT loss as a regularization term into the DPO loss to enhance training stability and generation fidelity. Additionally, by combining the FSDP framework with multiple memory optimization techniques, our approach achieves nearly three times higher training capacity than using FSDP alone. Extensive experiments on both I2V and T2V tasks across multiple datasets demonstrate that our method consistently outperforms existing approaches, delivering superior video generation quality.

cs.CV↗
arXiv2025

The $q$-Schur algebras in type $D$, I: fundamental multiplication formulas

By embedding the Hecke algebra $\check H_q$ of type $D$ into the Hecke algebra $H_{q,1}$ of type $B$ with unequal parameters $(q,1)$, the $q$-Schur algebras $S^κ_q(n,r)$ of type $D$ is naturally defined as the endomorphism algebra of the tensor space with the $\check H_q$-action restricted from the $H_{q,1}$-action that defines the $(q,1)$-Schur algebra $S^\jmath_{q,1}(n,r)$ of type $B$. We investigate the algebras $S^\jmath_{q,1}(n,r)$ and $S^κ_q(n,r)$ both algebraically and geometrically and describe their standard bases, dimension formulas and weight idempotents. Most importantly, we use the geometrically derived two sets of the fundamental multiplication formulas in $S^\jmath_{q,1}(n,r)$ to derive multi-sets (9 sets in total!) of the fundamental multiplication formulas in $S^κ_q(n,r)$.

math.QA↗
arXiv2024

Constructing the quantum queer supergroup using Hecke-Clifford superalgebras

In [DGLW], we use certain special elements and their commutation relations in the Hecke-Clifford algebras $H^c_{r,R}$ to derive some fundamental multiplication formulas associated with the natural bases in queer $q$-Schur superalgebras $Q_q(n,r;R)$ introduced in [DW2]. Here a natural basis element is defined by a special element $T_{A^{\star}}$ in $H^c_{r,R}$ associated with a pair of certain $n\times n$ matrices $A^{\star}=(A^{\bar0}|A^{\bar1})$ over $\mathbb{N}$ with entries sum to $r$. The definition of $T_{A^\star}$ consists of an element $c_{A^{\star}}$ in the Clifford superalgebra and an element $T_A$ in the Hecke algebra, where $A=A^{\bar0}+A^{\bar1}$. Note that all $T_A$ can be used to define the natural basis for the corresponding $q$-Schur algebra $S_q(n,r)$. This paper is a continuation of [DGLW]. We start with standardized queer $v$-Schur superalgebras $ Q^s_v(n,r)$, for $R=\mathbb{Z}[v,v^{-1}]$ and $q=v^2$, and their natural bases. With the $v$-Schur algebra ${ S}_v(n,r)$ at the background, the first key ingredient is a standardisation of the natural basis for $Q^s_v(n,r)$ and their associated standard multiplication formulas. By introducing some long elements of finite sums, we then extend the formulas to these long elements which allow us to explicitly define $\mathbb{Q}(v)$-superalgebra homomorphisms $ξ_{n,r}$ from the quantum queer supergroup $\boldsymbol{U}_v(\mathfrak{q}_n)$ to queer $q$-Schur superalgebras $\boldsymbol{Q}^s_v(n,r)$, for all $r\geq1$. Finally, taking limits of long elements yields certain infinitely long elements as formal infinite series which eventually lead to a new construction for $\boldsymbol{U}_v(\mathfrak{q}_n)$.

math.QA↗
arXiv2023

Approaching quantum queer supergroups using finite dimensional superalgebras (Preliminary version)

The idea of using a sequence of finite dimensional algebras to approach a quantum linear group (i.e., a quantum $\mathfrak{gl}_n$) was first introduced by Beilinson-Lusztig-MacPherson [BLM]. In their work, the algebras are convolution algebras of some finite partial flag varieties whose certain structure constants relative to the orbital basis satisfy a stabilization property. This property leads to the definition of an infinite dimensional idempotented algebra. Finally, taking a limit process yields a new realization for the quantum $\mathfrak{gl}_n$. Since then, this work has been modified [DF2] and generalized to quantum affine $\mathfrak{gl}_n$ (see [GV, L] for the geometric approach and [DDF, DF] for the algebraic approach and a new realization) and quantum super $\mathfrak{gl}_{m|n}$ [DG], and, more recently, to convolution algebras arising from type $B/C$ geometry and $i$-quantum groups $\boldsymbol U^\jmath$ and $\boldsymbol U^\imath$; see [BKLW, DWu1, DWu2]. This paper extends the algebraic approach to the quantum queer supergroup $U_{v}(\mathfrak{q}_n)$ via finite dimensional queer $q$-Schur superalgebras.

math.QA↗
arXiv2023

Some multiplication formulas in queer $q$-Schur superalgebras

Building on the work [18], where some standard basis for the queer $q$-Schur superalgebra $\mathcal{Q}_q(n,r;R)$ is defined by a labelling set of matrices and their associated double coset representatives, we investigate the matrix representation of the regular module of $\mathcal{Q}_q(n,r;R)$ with respect to this basis. More precisely, we derive explicitly (resp., partial explicitly) the multiplication formulas of the basis elements by certain even (resp., odd) generators of a queer $q$-Schur superalgebra. These multiplication formulas are highly technical to derive, especially in the odd case. It requires to discover many multiplication (or commutation) formulas in the Hecke--Clifford algebra $\mathcal{H}_{r,R}^c$ associated with the labelling matrices. For example, for a given such a labelling matrix $A^{\!\star}$, there are several matrices $w(A)$, $σ(A), \widetilde A$, and $\widehat A$ associated with the base matrix $A$ of $A^{\!\star}$, where $w(A)$ is used to compute a reduced expression of the distinguished double coset representatives $d_A$, and the other matrices are used to describe the permutation $d_A$ and the SDP (commutation) condition between $T_{d_A}$ and generators of the Clifford subsuperalgebra. With these multiplication formulas, we will construct a new realisation of the quantum queer supergroup in a forthcoming paper [13], and to give new applications to the integral Schur--Olshanski duality and its associated representation theory at roots of unity.

math.RT↗
arXiv2023

An exact category approach to Hecke endomorphism algebras

Let $G$ be a finite group of Lie type. In studying the cross-characteristic representation theory of $G$, the (specialized) Hecke algebra $H=\End_G(\ind_B^G1_B)$ has played a important role. In particular, when $G=GL_n(\mathbb F_q)$ is a finite general linear group, this approach led to the Dipper-James theory of $q$-Schur algebras $A$. These algebras can be constructed over $\sZ:=\mathbb Z[t,t^{-1}]$ as the $q$-analog (with $q=t^2$) of an endomorphism algebra larger than $H$, involving parabolic subgroups. The algebra $A$ is quasi-hereditary over $\sZ$. An analogous algebra, still denoted $A$, can always be constructed in other types. However, these algebras have so far been less useful than in the $GL_n$ case, in part because they are not generally quasi-hereditary. Several years ago, reformulating a 1998 conjecture, the authors proposed (for all types) the existence of a $\sZ$-algebra $A^+$ having a stratified derived module category, with strata constructed via Kazhdan-Lusztig cell theory. The algebra $A$ is recovered as $A=eA^+e$ for an idempotent $e\in A^+$. A main goal of this monograph is to prove this conjecture completely. The proof involves several new homological techniques using exact categories. Following the proof, we show that $A^+$ does become quasi-hereditary after the inversion of the bad primes. Some first applications of the result -- e.g., to decomposition matrices -- are presented, together with several open problems.

math.RT↗
arXiv2022

Stable Black Hole with Yang-Mills Hair

We present stable solution of static spherically symmetric Einstein-Yang-Mills equations with the SU(2) gauge group. This solution is asymptotically flat and regular at r = 0 and with nontrivial Yang-Mills(YM) connection. With quantized values of the Arnowitt-Deser-Misner (ADM) mass, the solutions asymptotically approach the Schwarzschild solution and have zero global YM charges. Numerical evidences suggest that this solution is both linearly and nonlinearly stable and has a ring of generic curvature singularities along the horizon. An effective counterexample to the no-hair conjecture is provided by this stable solution. Moreover, the stable black hole solution suggests that the coupling of gauge field to gravity in early Universe will generate a new type of black holes. Their stability means that these might be a possible new source of primordial black holes left over from the early Universe and serves as a possible new candidate for dark matter.

gr-qc↗
arXiv2022

Unsupervised Denoising of Optical Coherence Tomography Images with Dual_Merged CycleWGAN

Nosie is an important cause of low quality Optical coherence tomography (OCT) image. The neural network model based on Convolutional neural networks(CNNs) has demonstrated its excellent performance in image denoising. However, OCT image denoising still faces great challenges because many previous neural network algorithms required a large number of labeled data, which might cost much time or is expensive. Besides, these CNN-based algorithms need numerous parameters and good tuning techniques, which is hardware resources consuming. To solved above problems, We proposed a new Cycle-Consistent Generative Adversarial Nets called Dual-Merged Cycle-WGAN for retinal OCT image denoiseing, which has remarkable performance with less unlabeled traning data. Our model consists of two Cycle-GAN networks with imporved generator, descriminator and wasserstein loss to achieve good training stability and better performance. Using image merge technique between two Cycle-GAN networks, our model could obtain more detailed information and hence better training effect. The effectiveness and generality of our proposed network has been proved via ablation experiments and comparative experiments. Compared with other state-of-the-art methods, our unsupervised method obtains best subjective visual effect and higher evaluation objective indicators.

eess.IV↗
arXiv2021

Make A Long Image Short: Adaptive Token Length for Vision Transformers

The vision transformer splits each image into a sequence of tokens with fixed length and processes the tokens in the same way as words in natural language processing. More tokens normally lead to better performance but considerably increased computational cost. Motivated by the proverb "A picture is worth a thousand words" we aim to accelerate the ViT model by making a long image short. To this end, we propose a novel approach to assign token length adaptively during inference. Specifically, we first train a ViT model, called Resizable-ViT (ReViT), that can process any given input with diverse token lengths. Then, we retrieve the "token-length label" from ReViT and use it to train a lightweight Token-Length Assigner (TLA). The token-length labels are the smallest number of tokens to split an image that the ReViT can make the correct prediction, and TLA is learned to allocate the optimal token length based on these labels. The TLA enables the ReViT to process the image with the minimum sufficient number of tokens during inference. Thus, the inference speed is boosted by reducing the token numbers in the ViT model. Our approach is general and compatible with modern vision transformer architectures and can significantly reduce computational expanse. We verified the effectiveness of our methods on multiple representative ViT models (DeiT, LV-ViT, and TimesFormer) across two tasks (image classification and action recognition).

cs.CV↗
arXiv2021

Training BatchNorm Only in Neural Architecture Search and Beyond

This work investigates the usage of batch normalization in neural architecture search (NAS). Specifically, Frankle et al. find that training BatchNorm only can achieve nontrivial performance. Furthermore, Chen et al. claim that training BatchNorm only can speed up the training of the one-shot NAS supernet over ten times. Critically, there is no effort to understand 1) why training BatchNorm only can find the perform-well architectures with the reduced supernet-training time, and 2) what is the difference between the train-BN-only supernet and the standard-train supernet. We begin by showing that the train-BN-only networks converge to the neural tangent kernel regime, obtain the same training dynamics as train all parameters theoretically. Our proof supports the claim to train BatchNorm only on supernet with less training time. Then, we empirically disclose that train-BN-only supernet provides an advantage on convolutions over other operators, cause unfair competition between architectures. This is due to only the convolution operator being attached with BatchNorm. Through experiments, we show that such unfairness makes the search algorithm prone to select models with convolutions. To solve this issue, we introduce fairness in the search space by placing a BatchNorm layer on every operator. However, we observe that the performance predictor in Chen et al. is inapplicable on the new search space. To this end, we propose a novel composite performance indicator to evaluate networks from three perspectives: expressivity, trainability, and uncertainty, derived from the theoretical property of BatchNorm. We demonstrate the effectiveness of our approach on multiple NAS-benchmarks (NAS-Bench101, NAS-Bench-201) and search spaces (DARTS search space and MobileNet search space).

cs.LG↗
arXiv2021

The i-quantum group U^i(n)

This paper reveals some new structural property for the $i$-quantum group U^i(n) and constructs a certain hyperalgebra from the new structure which has connections to finite symplectic groups at the modular representation level.

math.QA↗
arXiv2020

A new realization of the i-quantum group U^j(n)

We follow the approach developed by Beilinson-Lusztig-MacPherson and modified by Fu and the first author to investigate a new realization for the i-quantum groups U^j(n) of type B, building on the multiplication formulas discovered in [BKLW,Lem.~3.2]. This allows us to present U^j(n) via a basis and multiplication formulas by generators. We also establish a surjective algebra homomorphism from a Lusztig type form of U^j(n) to integral q-Schur algebras of type B. Thus, base changes allow us to relate representations of the i-quantum hyperalgebras of U^j(n) to representations of finite orthogonal groups of odd degree in non-defining characteristics. This generalizes part of Dipper--James' type A theory to the type B case.

math.QA↗
arXiv2020

Quantum queer supergroups via v-differential operators

By using certain quantum differential operators, we construct a super representation for the quantum queer supergroup U_v(q_n). The underlying space of this representation is a deformed polynomial superalgebra in 2n^2 variables whose homogeneous components can be used as the underlying spaces of queer q-Schur superalgebras. We then extend the representation to its formal power series algebra which contains a (super) submodule isomorphic to the regular representation of U_v(q_n). A monomial basis M for U_v(q_n) plays a key role in proving the isomorphism. In this way, we may present the quantum queer supergroup U_v(q_n) by another new basis L together with some explicit multiplication formulas by the generators. As an application, similar presentations are obtained for queer q-Schur superalgebras via the above mentioned homogeneous components. The existence of the bases M and L and the new presentation show that the seminal construction of quantum gl_n established by Beilinson-Lusztig-MacPherson thirty years ago extends to this "queer" quantum supergroup via a completely different approach.

math.QA↗