Author name cluster

Xiang He

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

10 papers

2 author rows

AAAI Conference 2026 Conference Paper

Plug-and-Play Parameter-Efficient Tuning of Embeddings for Federated Recommendation

Haochen Yuan
Yang Zhang
Xiang He
Quan Z. Sheng
Zhongjie Wang

With the rise of cloud-edge collaboration, recommendation services are increasingly trained in distributed environments. Federated Recommendation (FR) enables such multi-end collaborative training while preserving privacy by sharing model parameters instead of raw data. However, the large number of parameters, primarily due to the massive item embeddings, significantly hampers communication efficiency. While existing studies mainly focus on improving the efficiency of FR models, they largely overlook the issue of embedding parameter overhead. To address this gap, we propose a FR training framework with Parameter-Efficient Fine-Tuning (PEFT) based embedding designed to reduce the volume of embedding parameters that need to be transmitted. Our approach offers a lightweight, plugin-style solution that can be seamlessly integrated into existing FR methods. In addition to incorporating common PEFT techniques such as LoRA and Hash-based encoding, we explore the use of Residual Quantized Variational Autoencoders (RQ-VAE) as a novel PEFT strategy within our framework. Extensive experiments across various FR model backbones and datasets demonstrate that our framework significantly reduces communication overhead while improving accuracy.

PDF Details DOI

IROS Conference 2025 Conference Paper

A Dual Tiltrotor UAV with Foldable Wings for Passive Perching and Belly/Back Takeoff

Luyao Wang
Jiangyi Zhang
Liangliang Cheng
Jingrui Yang
Tianchi Ma
Xiang He

This paper presents a novel dual tiltrotor UAV design featuring foldable wings and a strategically positioned center of gravity (CG) to enable passive perching and multi-modal flight. Traditional UAVs rely on additional mechanical components for operations such as takeoff and perching, which increase weight and complexity. Inspired by the mechanics of a balanced bird toy, our design achieves stability in horizontal flight and secure power-off perching on branches or cables. The proposed blade-tip plane-based controller facilitates belly/back takeoff without landing gear, enabling seamless transitions between hovering and horizontal flight within 2 seconds. Wind resistance tests were conducted indoors to assess disturbance rejection capabilities during perching, while the transition performance was evaluated outdoors.

Details

AAAI Conference 2025 Conference Paper

EventZoom: A Progressive Approach to Event-Based Data Augmentation for Enhanced Neuromorphic Vision

Yiting Dong
Xiang He
Guobin Shen
Dongcheng Zhao
Yang Li
Yi Zeng

Dynamic Vision Sensors (DVS) capture event data with high temporal resolution and low power consumption, presenting a more efficient solution for visual processing in dynamic and real-time scenarios compared to conventional video capture methods. Event data augmentation serves as an essential method for overcoming the limitation of scale and diversity in event datasets. Our comparative experiments demonstrate that the two factors, spatial integrity and temporal continuity, can significantly affect the capacity of event data augmentation, which guarantee the maintenance of the sparsity and high dynamic range characteristics unique to event data. However, existing augmentation methods often neglect the preservation of spatial integrity and temporal continuity. To address this, we developed a novel event data augmentation strategy EventZoom, which employs a temporal progressive strategy, embedding transformed samples into the original samples through progressive scaling and shifting. The scaling process avoids the spatial information loss associated with cropping, while the progressive strategy prevents interruptions or abrupt changes in temporal information. We validated EventZoom across various supervised learning frameworks. The experimental results show that EventZoom consistently outperforms existing event data augmentation methods with SOTA performance. For the first time, we have concurrently employed Semi-supervised and Unsupervised learning to verify feasibility on event augmentation algorithms, demonstrating the applicability and effectiveness of EventZoom as a powerful event-based data augmentation tool in handling real-world scenes with high dynamics and variability environments.

PDF Details DOI

AAAI Conference 2025 Conference Paper

StressPrompt: Does Stress Impact Large Language Models and Human Performance Similarly?

Guobin Shen
Dongcheng Zhao
Aorigele Bao
Xiang He
Yiting Dong
Yi Zeng

Human beings often experience stress, which can significantly influence their performance. This study explores whether Large Language Models (LLMs) exhibit stress responses similar to those of humans and whether their performance fluctuates under different stress-inducing prompts. To investigate this, we developed a novel set of prompts, termed StressPrompt, designed to induce varying levels of stress. These prompts were derived from established psychological frameworks and carefully calibrated based on ratings from human participants. We then applied these prompts to several LLMs to assess their responses across a range of tasks, including instruction-following, complex reasoning, and emotional intelligence. The findings suggest that LLMs, like humans, perform optimally under moderate stress, consistent with the Yerkes-Dodson law. Notably, their performance declines under both low and high-stress conditions. Our analysis further revealed that these StressPrompts significantly alter the internal states of LLMs, leading to changes in their neural representations that mirror human responses to stress. This research provides critical insights into the operational robustness and flexibility of LLMs, demonstrating the importance of designing AI systems capable of maintaining high performance in real-world scenarios where stress is prevalent, such as in customer service, healthcare, and emergency response contexts. Moreover, this study contributes to the broader AI research community by offering a new perspective on how LLMs handle different scenarios and their similarities to human cognition.

PDF Details DOI

AAAI Conference 2024 Conference Paper

An Efficient Knowledge Transfer Strategy for Spiking Neural Networks from Static to Event Domain

Xiang He
Dongcheng Zhao
Yang Li
Guobin Shen
Qingqun Kong
Yi Zeng

Spiking neural networks (SNNs) are rich in spatio-temporal dynamics and are suitable for processing event-based neuromorphic data. However, event-based datasets are usually less annotated than static datasets. This small data scale makes SNNs prone to overfitting and limits their performance. In order to improve the generalization ability of SNNs on event-based datasets, we use static images to assist SNN training on event data. In this paper, we first discuss the domain mismatch problem encountered when directly transferring networks trained on static datasets to event data. We argue that the inconsistency of feature distributions becomes a major factor hindering the effective transfer of knowledge from static images to event data. To address this problem, we propose solutions in terms of two aspects: feature distribution and training strategy. Firstly, we propose a knowledge transfer loss, which consists of domain alignment loss and spatio-temporal regularization. The domain alignment loss learns domain-invariant spatial features by reducing the marginal distribution distance between the static image and the event data. Spatio-temporal regularization provides dynamically learnable coefficients for domain alignment loss by using the output features of the event data at each time step as a regularization term. In addition, we propose a sliding training strategy, which gradually replaces static image inputs probabilistically with event data, resulting in a smoother and more stable training for the network. We validate our method on neuromorphic datasets, including N-Caltech101, CEP-DVS, and N-Omniglot. The experimental results show that our proposed method achieves better performance on all datasets compared to the current state-of-the-art methods. Code is available at https://github.com/Brain-Cog-Lab/Transfer-for-DVS.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

Neuro-Vision to Language: Enhancing Brain Recording-based Visual Reconstruction and Language Interaction

Guobin Shen
Dongcheng Zhao
Xiang He
Linghao Feng
Yiting Dong
Jihang Wang
Qian Zhang
Yi Zeng

Decoding non-invasive brain recordings is pivotal for advancing our understanding of human cognition but faces challenges due to individual differences and complex neural signal representations. Traditional methods often require customized models and extensive trials, lacking interpretability in visual reconstruction tasks. Our framework integrates 3D brain structures with visual semantics using a Vision Transformer 3D. This unified feature extractor efficiently aligns fMRI features with multiple levels of visual embeddings, eliminating the need for subject-specific models and allowing extraction from single-trial data. The extractor consolidates multi-level visual features into one network, simplifying integration with Large Language Models (LLMs). Additionally, we have enhanced the fMRI dataset with diverse fMRI-image-related textual data to support multimodal large model development. Integrating with LLMs enhances decoding capabilities, enabling tasks such as brain captioning, complex reasoning, concept localization, and visual reconstruction. Our approach demonstrates superior performance across these tasks, precisely identifying language-based concepts within brain signals, enhancing interpretability, and providing deeper insights into neural processes. These advances significantly broaden the applicability of non-invasive brain decoding in neuroscience and human-computer interaction, setting the stage for advanced brain-computer interfaces and cognitive models.

PDF Details DOI

YNIMG Journal 2020 Journal Article

Reproducibility assessment of neuromelanin-sensitive magnetic resonance imaging protocols for region-of-interest and voxelwise analyses

Kenneth Wengler
Xiang He
Anissa Abi-Dargham
Guillermo Horga

Neuromelanin-sensitive MRI (NM-MRI) provides a noninvasive measure of the content of neuromelanin (NM), a product of dopamine metabolism that accumulates with age in dopamine neurons of the substantia nigra (SN). NM-MRI has been validated as a measure of both dopamine neuron loss, with applications in neurodegenerative disease, and dopamine function, with applications in psychiatric disease. Furthermore, a voxelwise-analysis approach has been validated to resolve substructures, such as the ventral tegmental area (VTA), within midbrain dopaminergic nuclei thought to have distinct anatomical targets and functional roles. NM-MRI is thus a promising tool that could have diverse research and clinical applications to noninvasively interrogate in vivo the dopamine system in neuropsychiatric illness. Although a test-retest reliability study by Langley et al. using the standard NM-MRI protocol recently reported high reliability, a systematic and comprehensive investigation of the performance of the method for various acquisition parameters and preprocessing methods has not been conducted. In particular, most previous studies used relatively thick MRI slices (~3 mm), compared to the typical in-plane resolution (~0. 5 mm) and to the height of the SN (~15 mm), to overcome technical limitations such as specific absorption rate and signal-to-noise ratio, at the cost of partial-volume effects. Here, we evaluated the effect of various acquisition and preprocessing parameters on the strength and test-retest reliability of the NM-MRI signal to determine optimized protocols for both region-of-interest (including whole SN-VTA complex and atlas-defined dopaminergic nuclei) and voxelwise measures. Namely, we determined a combination of parameters that optimizes the strength and reliability of the NM-MRI signal, including acquisition time, slice-thickness, spatial-normalization software, and degree of spatial smoothing. Using a newly developed, detailed acquisition protocol, across two scans separated by 13 days on average, we obtained intra-class correlation values indicating excellent reliability and high contrast, which could be achieved with a different set of parameters depending on the measures of interest and experimental constraints such as acquisition time. Based on this, we provide detailed guidelines covering acquisition through analysis and recommendations for performing NM-MRI experiments with high quality and reproducibility. This work provides a foundation for the optimization and standardization of NM-MRI, a promising MRI approach with growing applications throughout clinical and basic neuroscience.

Details DOI

YNIMG Journal 2019 Journal Article

3D MRI of whole-brain water permeability with intrinsic diffusivity encoding of arterial labeled spin (IDEALS)

Kenneth Wengler
Lev Bangiyev
Turhan Canli
Tim Q. Duong
Mark E. Schweitzer
Xiang He

This work proposes a novel MRI method – Intrinsic Diffusivity Encoding of Arterial Labeled Spin (IDEALS) – for the whole-brain mapping of water permeability in the human brain without an exogenous contrast agent. Quantitative separation of the intravascular and extravascular labeled water MRI signal was achieved in arterial spin labeling experiments with segmented 3D-GRASE acquisition by modulating the relative sensitivity between relaxation, true diffusion, and pseudodiffusion. The intrinsic diffusivity encoding in k-space created different broadening of the image-domain point spread functions for intravascular and extravascular labeled spins, from which blood-brain barrier (BBB) water extraction fraction (E w ) and water permeability surface area product (PS w ) were estimated. The feasibility and sensitivity of this method was evaluated in healthy subjects at baseline and after caffeine challenge. The estimated baseline E w and PS w maps showed contrast among gray matter (GM) and white matter (WM). GM E w was significantly lower than that of WM (78. 8% ± 3. 3% in GM vs. 83. 9% ± 4. 6% in WM; p < 0. 05) and GM PS w was significantly higher than that of WM (131. 7 ± 29. 5 mL/100 g/min in GM vs. 76. 2 ± 18. 4 mL/100 g/min in WM; p < 0. 05). BBB E w was significantly lower for females than males (74. 9% ± 3. 7% for females vs. 81. 3% ± 3. 3% for males in GM; 80. 5% ± 4. 7% for females vs. 86. 1 ± 3. 0 for males in WM; p < 0. 05 for both), while significant PS w differences were only observed in WM (143. 8 ± 34. 4 mL/100 g/min for females vs. 123. 6 ± 24. 4 mL/100 g/min for males in GM; 91. 6 ± 15. 0 mL/100 g/min for females vs. 65. 9 ± 12. 5 mL/100 g/min for males in WM; p = 0. 20 and p < 0. 05 for GM and WM respectively). Significant correlations between E w and CBF (r = −0. 32, p < 0. 05) and between PS w and CBF (r = 0. 89, p < 0. 05) were observed, consistent with 15O-H2O PET findings. After caffeine challenge, reduced CBF, E w and PS w were observed, demonstrating the sensitivity of IDEALS approach.

Details DOI

AAAI Conference 2019 Conference Paper

Non-Local Context Encoder: Robust Biomedical Image Segmentation against Adversarial Attacks

Xiang He
Sibei Yang
Guanbin Li
Haofeng Li
Huiyou Chang
Yizhou Yu

Recent progress in biomedical image segmentation based on deep convolutional neural networks (CNNs) has drawn much attention. However, its vulnerability towards adversarial samples cannot be overlooked. This paper is the first one that discovers that all the CNN-based state-of-the-art biomedical image segmentation models are sensitive to adversarial perturbations. This limits the deployment of these methods in safety-critical biomedical fields. In this paper, we discover that global spatial dependencies and global contextual information in a biomedical image can be exploited to defend against adversarial attacks. To this end, non-local context encoder (NLCE) is proposed to model short- and longrange spatial dependencies and encode global contexts for strengthening feature activations by channel-wise attention. The NLCE modules enhance the robustness and accuracy of the non-local context encoding network (NLCEN), which learns robust enhanced pyramid feature representations with NLCE modules, and then integrates the information across different levels. Experiments on both lung and skin lesion segmentation datasets have demonstrated that NLCEN outperforms any other state-of-the-art biomedical image segmentation methods against adversarial attacks. In addition, NLCE modules can be applied to improve the robustness of other CNN-based biomedical image segmentation methods.

PDF Details

IROS Conference 2002 Conference Paper

Dense depth map acquisition by hierarchic structured light

Peiyi Niu
Xiang He
Andrew K. C. Wong

Structured light systems (SLS) have been used in 3D model building for a long time. However, the comprehensive performance of such systems, in terms of data acquisition speed and depth density, is still far from being perfect. In this paper, we introduce a new pattern design for SLS in order to get a dense depth map while in principle maintaining the acquisition speed, high reliability and accuracy. Our system uses a hierarchical integration of features and textures in the projection pattern. This hierarchy is then used to guide the matching process. The effectiveness of the method has been demonstrated by extensive experiments.

Details