arXiv subjects

Xinyu Du

Publications and source records attributed to Xinyu Du.

17 recordsLinked to original sources

arXiv2025

Rhetorical Text-to-Image Generation via Two-layer Diffusion Policy Optimization

Generating images from rhetorical languages remains a critical challenge for text-to-image models. Even state-of-the-art (SOTA) multimodal large language models (MLLM) fail to generate images based on the hidden meaning inherent in rhetorical language--despite such content being readily mappable to visual representations by humans. A key limitation is that current models emphasize object-level word embedding alignment, causing metaphorical expressions to steer image generation towards their literal visuals and overlook the intended semantic meaning. To address this, we propose Rhet2Pix, a framework that formulates rhetorical text-to-image generation as a multi-step policy optimization problem, incorporating a two-layer MDP diffusion module. In the outer layer, Rhet2Pix converts the input prompt into incrementally elaborated sub-sentences and executes corresponding image-generation actions, constructing semantically richer visuals. In the inner layer, Rhet2Pix mitigates reward sparsity during image generation by discounting the final reward and optimizing every adjacent action pair along the diffusion denoising trajectory. Extensive experiments demonstrate the effectiveness of Rhet2Pix in rhetorical text-to-image generation. Our model outperforms SOTA MLLMs such as GPT-4o, Grok-3 and leading academic baselines across both qualitative and quantitative evaluations. The code and dataset used in this work are publicly available.

cs.CV↗

arXiv2025

MDD-LLM: Towards Accuracy Large Language Models for Major Depressive Disorder Diagnosis

Major depressive disorder (MDD) impacts more than 300 million people worldwide, highlighting a significant public health issue. However, the uneven distribution of medical resources and the complexity of diagnostic methods have resulted in inadequate attention to this disorder in numerous countries and regions. This paper introduces a high-performance MDD diagnosis tool named MDD-LLM, an AI-driven framework that utilizes fine-tuned large language models (LLMs) and extensive real-world samples to tackle challenges in MDD diagnosis. Therefore, we select 274,348 individual information from the UK Biobank cohort to train and evaluate the proposed method. Specifically, we select 274,348 individual records from the UK Biobank cohort and design a tabular data transformation method to create a large corpus for training and evaluating the proposed approach. To illustrate the advantages of MDD-LLM, we perform comprehensive experiments and provide several comparative analyses against existing model-based solutions across multiple evaluation metrics. Experimental results show that MDD-LLM (70B) achieves an accuracy of 0.8378 and an AUC of 0.8919 (95% CI: 0.8799 - 0.9040), significantly outperforming existing machine learning and deep learning frameworks for MDD diagnosis. Given the limited exploration of LLMs in MDD diagnosis, we examine numerous factors that may influence the performance of our proposed method, such as tabular data transformation techniques and different fine-tuning strategies.

cs.CL↗

arXiv2025

Detecting Credit Card Fraud via Heterogeneous Graph Neural Networks with Graph Attention

This study proposes a credit card fraud detection method based on Heterogeneous Graph Neural Network (HGNN) to address fraud in complex transaction networks. Unlike traditional machine learning methods that rely solely on numerical features of transaction records, this approach constructs heterogeneous transaction graphs. These graphs incorporate multiple node types, including users, merchants, and transactions. By leveraging graph neural networks, the model captures higher-order transaction relationships. A Graph Attention Mechanism is employed to dynamically assign weights to different transaction relationships. Additionally, a Temporal Decay Mechanism is integrated to enhance the model's sensitivity to time-related fraud patterns. To address the scarcity of fraudulent transaction samples, this study applies SMOTE oversampling and Cost-sensitive Learning. These techniques strengthen the model's ability to identify fraudulent transactions. Experimental results demonstrate that the proposed method outperforms existing GNN models, including GCN, GAT, and GraphSAGE, on the IEEE-CIS Fraud Detection dataset. The model achieves notable improvements in both accuracy and OC-ROC. Future research may explore the integration of dynamic graph neural networks and reinforcement learning. Such advancements could enhance the real-time adaptability of fraud detection systems and provide more intelligent solutions for financial risk control.

cs.LG↗

arXiv2024

Semantic Feature Learning for Universal Unsupervised Cross-Domain Retrieval

Cross-domain retrieval (CDR), as a crucial tool for numerous technologies, is finding increasingly broad applications. However, existing efforts face several major issues, with the most critical being the need for accurate supervision, which often demands costly resources and efforts. Cutting-edge studies focus on achieving unsupervised CDR but typically assume that the category spaces across domains are identical, an assumption that is often unrealistic in real-world scenarios. This is because only through dedicated and comprehensive analysis can the category spaces of different domains be confirmed as identical, which contradicts the premise of unsupervised scenarios. Therefore, in this work, we introduce the problem of Universal Unsupervised Cross-Domain Retrieval (U^2CDR) for the first time and design a two-stage semantic feature learning framework to address it. In the first stage, a cross-domain unified prototypical structure is established under the guidance of an instance-prototype-mixed contrastive loss and a semantic-enhanced loss, to counteract category space differences. In the second stage, through a modified adversarial training mechanism, we ensure minimal changes for the established prototypical structure during domain alignment, enabling more accurate nearest-neighbor searching. Extensive experiments across multiple datasets and scenarios, including closet, partial, and open-set CDR, demonstrate that our approach significantly outperforms existing state-of-the-art CDR works and some potentially effective studies from other topics in solving U^2CDR challenges.

cs.CV↗

arXiv2024

DACR: Distribution-Augmented Contrastive Reconstruction for Time-Series Anomaly Detection

Anomaly detection in time-series data is crucial for identifying faults, failures, threats, and outliers across a range of applications. Recently, deep learning techniques have been applied to this topic, but they often struggle in real-world scenarios that are complex and highly dynamic, e.g., the normal data may consist of multiple distributions, and various types of anomalies may differ from the normal data to different degrees. In this work, to tackle these challenges, we propose Distribution-Augmented Contrastive Reconstruction (DACR). DACR generates extra data disjoint from the normal data distribution to compress the normal data's representation space, and enhances the feature extractor through contrastive learning to better capture the intrinsic semantics from time-series data. Furthermore, DACR employs an attention mechanism to model the semantic dependencies among multivariate time-series features, thereby achieving more robust reconstruction for anomaly detection. Extensive experiments conducted on nine benchmark datasets in various anomaly detection scenarios demonstrate the effectiveness of DACR in achieving new state-of-the-art time-series anomaly detection.

cs.LG↗

arXiv2023

Frequency Enhanced Hybrid Attention Network for Sequential Recommendation

The self-attention mechanism, which equips with a strong capability of modeling long-range dependencies, is one of the extensively used techniques in the sequential recommendation field. However, many recent studies represent that current self-attention based models are low-pass filters and are inadequate to capture high-frequency information. Furthermore, since the items in the user behaviors are intertwined with each other, these models are incomplete to distinguish the inherent periodicity obscured in the time domain. In this work, we shift the perspective to the frequency domain, and propose a novel Frequency Enhanced Hybrid Attention Network for Sequential Recommendation, namely FEARec. In this model, we firstly improve the original time domain self-attention in the frequency domain with a ramp structure to make both low-frequency and high-frequency information could be explicitly learned in our approach. Moreover, we additionally design a similar attention mechanism via auto-correlation in the frequency domain to capture the periodic characteristics and fuse the time and frequency level attention in a union model. Finally, both contrastive learning and frequency regularization are utilized to ensure that multiple views are aligned in both the time domain and frequency domain. Extensive experiments conducted on four widely used benchmark datasets demonstrate that the proposed model performs significantly better than the state-of-the-art approaches.

cs.IR↗

arXiv2023

Contrastive Enhanced Slide Filter Mixer for Sequential Recommendation

Sequential recommendation (SR) aims to model user preferences by capturing behavior patterns from their item historical interaction data. Most existing methods model user preference in the time domain, omitting the fact that users' behaviors are also influenced by various frequency patterns that are difficult to separate in the entangled chronological items. However, few attempts have been made to train SR in the frequency domain, and it is still unclear how to use the frequency components to learn an appropriate representation for the user. To solve this problem, we shift the viewpoint to the frequency domain and propose a novel Contrastive Enhanced \textbf{SLI}de Filter \textbf{M}ixEr for Sequential \textbf{Rec}ommendation, named \textbf{SLIME4Rec}. Specifically, we design a frequency ramp structure to allow the learnable filter slide on the frequency spectrums across different layers to capture different frequency patterns. Moreover, a Dynamic Frequency Selection (DFS) and a Static Frequency Split (SFS) module are proposed to replace the self-attention module for effectively extracting frequency information in two ways. DFS is used to select helpful frequency components dynamically, and SFS is combined with the dynamic frequency selection module to provide a more fine-grained frequency division. Finally, contrastive learning is utilized to improve the quality of user embedding learned from the frequency domain. Extensive experiments conducted on five widely used benchmark datasets demonstrate our proposed model performs significantly better than the state-of-the-art approaches. Our code is available at https://github.com/sudaada/SLIME4Rec.

cs.IR↗

arXiv2023

Online Camera-to-ground Calibration for Autonomous Driving

Online camera-to-ground calibration is to generate a non-rigid body transformation between the camera and the road surface in a real-time manner. Existing solutions utilize static calibration, suffering from environmental variations such as tire pressure changes, vehicle loading volume variations, and road surface diversity. Other online solutions exploit the usage of road elements or photometric consistency between overlapping views across images, which require continuous detection of specific targets on the road or assistance with multiple cameras to facilitate calibration. In our work, we propose an online monocular camera-to-ground calibration solution that does not utilize any specific targets while driving. We perform a coarse-to-fine approach for ground feature extraction through wheel odometry and estimate the camera-to-ground calibration parameters through a sliding-window-based factor graph optimization. Considering the non-rigid transformation of camera-to-ground while driving, we provide metrics to quantify calibration performance and stopping criteria to report/broadcast our satisfying calibration results. Extensive experiments using real-world data demonstrate that our algorithm is effective and outperforms state-of-the-art techniques.

cs.CV↗

arXiv2022

Deep Near-Infrared Survey Toward the W40 and Serpens South Region in Aquila Rift: A Comprehensive Catalog of Young Stellar Objects

Active star forming regions are excellent laboratories for studying the origins and evolution of young stellar object (YSO) clustering. The W40 - Serpens South region is such a region, and we compile a super near-and-mid-infrared catalog of point sources in it, based on deep NIR observations of CFHT in combination with 2MASS, UKIDSS, and Spitzer catalogs. From this catalog, we identify 832 YSOs, and classify 15, 135, 647, and 35 of them to be the deeply embedded sources, Class I, Class II YSOs, and transition disk sources, respectively. In general, these YSOs are well correlated with the filamentary structures of molecular clouds, especially the deeply embedded sources and the Class I YSOs. The W40 central region is dominated by Class II YSOs, but in the Serpens South region, a half of the YSOs are Class I. We further generate a minimum spanning tree (MST) for all the YSOs. Around the W40 cluster, there are eight prominent MST branches that may trace vestigial molecular gas filaments that once fed gas to the central natal gas clump. Of the eight, only two now include detectable filamentary gas in Herschel data and corresponding Class I YSOs, while the other six are exclusively populated with Class II. Four MST branches overlap with the Serpens South main filament, and where they intersect, molecular gas "hubs" and more Class I YSOs are found. Our results imply a mixture of YSO distributions composed of both primordial and somewhat evolved YSOs in this star forming region.

astro-ph.SR↗

arXiv2019

Efficient methods to determine the reversibility of general 1D linear cellular automata in polynomial complexity

In this paper, we study reversibility of one-dimensional(1D) linear cellular automata(LCA) under null boundary condition, whose core problems have been divided into two main parts: calculating the period of reversibility and verifying the reversibility in a period. With existing methods, the time and space complexity of these two parts are still too expensive to be employed. So the process soon becomes totally incalculable with a slightly big size, which greatly limits its application. In this paper, we set out to solve these two problems using two efficient algorithms, which make it possible to solve reversible LCA of very large size. Furthermore, we provide an interesting perspective to conversely generate 1D LCA from a given period of reversibility. Due to our methods' efficiency, we can calculate the reversible LCA with large size, which has much potential to enhance security in cryptography system.

cs.CC↗

arXiv2017

A 95 GHz Methanol Emission Survey Toward Eight Small Supernova Remnants

We report on a 95 GHz ($8_0-7_1$ A$^{+}$) methanol (CH$_3$OH) emission survey with the Purple Mountain Observatory Delingha 13.7 m telescope. Eight supernova remnants (SNRs) with angular size $\lesssim$ 10' were observed, but emission was only detected in three SNRs near the Galactic center (Sgr A East, G 0.1-0.1, and G 359.92-0.09). CH$_3$OH emission mainly surrounds the SNRs and can be decomposed into nine spatial peaks with velocity range of eight peaks being (-30, 70) km s$^{-1}$, and the other (70, 120) km s$^{-1}$. They are probably excited by interaction with these SNRs and adjacent molecular gas in the central molecular zone (CMZ), although star formation may play an important role in exciting CH$_3$OH emission in some regions of CMZ. We infer that tidal action is unlikely to be an excitation source for CH$_3$OH emission.

astro-ph.GA↗

arXiv2017

Observation of Unusual Magnetoelastic Effects in a Quasi-1D Spiral Magnet

We present a systematic study of spin and lattice dynamics in the quasi-one-dimensional spiral magnet CuBr2, using Raman scattering in conjunction with infrared and neutron spectroscopy. Along with the development of spin correlations upon cooling, we observe a rich set of broad Raman bands at energies that correspond to phonon-dispersion energies near the one-dimensional magnetic wave vector. The low-energy bands further exhibit a distinct intensity maximum at the spiral magnetic ordering temperature. We attribute these unusual observations to two possible underlying mechanisms: (1) formation of hybrid spin-lattice excitations, and/or (2) "quadrumerization" of the lattice caused by spin-singlet entanglement in competition with the spiral magnetism.

cond-mat.str-el↗

arXiv2017

The Molecular Structures of Local Arm and Perseus Arm in the Galactic Region of l=[139.75,149.75]$^\circ$, b=[-5.25,5.25]$^\circ$

Using the Purple Mountain Observatory Delingha (PMODLH) 13.7 m telescope, we report a 96-square-degree 12CO/13CO/C18O mapping observation toward the Galactic region of l = [139.75, 149.75]$^\circ$, b = [-5.25, 5.25]$^\circ$. The molecular structure of the Local Arm and Perseus Arm are presented. Combining HI data and part of the Outer Arm results, we obtain that the warp structure of both atomic and molecular gas is obvious, while the flare structure only exists in atomic gas in this observing region. In addition, five filamentary giant molecular clouds on the Perseus Arm are identified. Among them, four are newly identified. Their relations with the Milky Way large-scale structure are discussed.

astro-ph.GA↗

arXiv2017

CO~($J = 1-0$) Observations of a Filamentary Molecular Cloud in the Galactic Region Centered at $l = 150\arcdeg, b = 3.5\arcdeg$

We present large-field (4.25~$\times$~3.75 deg$^2$) mapping observations toward the Galactic region centered at $l = 150\arcdeg, b = 3.5\arcdeg$ in the $J = 1-0$ emission line of CO isotopologues ($^{12}$CO, $^{13}$CO, and C$^{18}$O), using the 13.7 m millimeter-wavelength telescope of the Purple Mountain Observatory. Based on the $^{13}$CO observations, we reveal a filamentary cloud in the Local Arm at a velocity range of $-$0.5 to 6.5~km~s$^{-1}$. This molecular cloud contains 1 main filament and 11 sub-filaments, showing the so-called "ridge-nest" structure. The main filament and three sub-filaments are also detected in the C$^{18}$O line. The velocity structures of most identified filaments display continuous distribution with slight velocity gradients. The measured median excitation temperature, line width, length, width, and linear mass of the filaments are $\sim$9.28~K, 0.85~km~s$^{-1}$, 7.30~pc, 0.79~pc, and 17.92~$M_\sun$~pc$^{-1}$, respectively, assuming a distance of 400~pc. We find that the four filaments detected in the C$^{18}$O line are thermally supercritical, and two of them are in the virialized state, and thus tend to be gravitationally bound. We identify in total 146 $^{13}$CO clumps in the cloud, about 77$\%$ of the clumps are distributed along the filaments. About 56$\%$ of the virialized clumps are found to be associated with the supercritical filaments. Three young stellar object (YSO) candidates are also identified in the supercritical filaments, based on the complementary infrared (IR) data. These results indicate that the supercritical filaments, especially the virialized filaments, may contain star-forming activities.

astro-ph.GA↗

arXiv2016

An Outer Arm in the Second Galactic Quadrant: Structure

The lack of arm tracers, especially the remote tracers, is one of the most difficult problems preventing us from studying the structure of the Milky Way. Fortunately, with its high-sensitivity CO survey, the Milky Way Imaging Scroll Painting (MWISP) project offers such an opportunity. Since completing about one-third of its mission, an area of l=[100,150] deg, b=[-3,5] deg has nearly been covered. The Outer arm of the Milky Way first clearly revealed its shape in the second galactic quadrant in the form of molecular gas --- this is the first time that the Outer arm has been reported in such a large-scale mapping of molecular gas. Using the 115 GHz 12CO(1-0) data of MWISP at the LSR velocity ~= [-100,-60] km s^-1 and in the area mentioned above, we have detected 481 molecular clouds in total, and among them 332 (about 69\%) are newly detected and 457 probably belong to the Outer arm. The total mass of the detected Outer arm clouds is ~ 3.1*10^6 M_sun. Assuming that the spiral arm is a logarithmic spiral, the pitch angle is fitted as ~ 13.1 deg. Besides combining both the CO data from MWISP and the 21 cm HI data from the Canadian Galactic Plane Survey (CGPS), the gas distribution, warp, and thickness of the Outer arm are also studied.

astro-ph.GA↗

arXiv2015

Identification and mechanical control of ferroelastic domain structure in rhombohedral CaMn$_7$O$_{12}$

We report on observation of ferroelastic domain structure in single crystals of multiferroic CaMn$_7$O$_{12}$ at room temperature. Two types of ferroelastic domain wall are found, consistent with the material's rhombohedral symmetry that is reduced from cubic symmetry at higher temperatures. Using Raman spectroscopy along with other measurements, we develop a systematic method to determine the microscopic domain orientation. Moreover, we find a switching behavior of the domains, which allows us to detwin the crystals conveniently at room temperature using a moderate uniaxial compression. Our result paves the way for further spectroscopic study and domain engineering in CaMn$_7$O$_{12}$.

cond-mat.str-el↗

arXiv2014

Soft vibrational mode associated with incommensurate orbital order in multiferroic CaMn$_7$O$_{12}$

We report inelastic light scattering measurements of lattice dynamics related to the incommensurate orbital order in $\mathrm{CaMn_7O_{12}}$. Below the ordering temperature $T_\mathrm{o} \approx 250 \,\mathrm{K}$, we observe extra phonon peaks as a result of Brillouin-zone folding, as well as a soft vibrational mode with a power-law $T$-dependent energy, $Ω= Ω_{0}(1 - T/T_{\mathrm{o}})^{1/2}$. This temperature dependence demonstrates the second-order nature of the transition at $T_\mathrm{o}$, and it indicates that the soft mode can be regarded as the amplitude excitation of the composite order parameter. Our result strongly suggests that the lattice degrees of freedom are actively involved in the orbital-ordering mechanism.

cond-mat.str-el↗