Huazhong University of Science and Technology
Ph.D. Candidate in Intelligent Science and Technology
I develop data-efficient, robust, and generalizable learning methods for neural signals under limited labels, strong heterogeneity, and distribution shifts. My research spans cross-species and cross-modality transfer, interpretable EEG decoding, multimodal learning, and physiologically grounded data generation.
Research vision
My long-term goal is to build trustworthy cross-species, multimodal, and multitask foundation models for brain-disorder screening, prognosis, treatment-response monitoring, and rehabilitation.
I am particularly interested in shared neural representations across epilepsy, Alzheimer’s disease, hypoxic–ischemic encephalopathy (HIE), amyotrophic lateral sclerosis (ALS), and related neurological disorders.
Research
My work connects transfer, decoding, and data generation into a unified path toward general-purpose brain foundation models.
Generalization
Transfer neural knowledge across species, modalities, devices, datasets, and patients when target labels are scarce.
Representation
Design EEG-specific architectures that explicitly model temporal dynamics, channel topology, frequency structure, and modality reliability.
Data foundation
Build physiologically grounded augmentation, generation, denoising, and evaluation methods for scalable neural-model pretraining.
Recent publications, open resources, funding, and honors. Drag or scroll →
DenoFlow was accepted by Knowledge-Based Systems.
Released the KDHNet preprint.
Awarded the Ph.D. National Scholarship (国家奖学金) for the third consecutive year, marking four National Scholarships overall.
Released preprints of CoMA-DiT and RAMamba-Net.
The NSR paper Canine EEG Helps Human received the Excellent Scientific and Technical Paper Award from the Hubei Association for Science and Technology (湖北省科协优秀科技论文奖).
CKD was accepted by IEEE Transactions on Biomedical Engineering.
Released the FAConformer preprint for frequency-aware auditory attention decoding.
Released the CHSZ pediatric epilepsy EEG dataset, collected with Wuhan Children’s Hospital.
Released DG4BCI, a unified survey and benchmark for synthetic data generation in BCIs.
The cross-species EEG study was selected for the Top 10 Advances in Brain–Computer Interfaces in China (中国脑机接口十大进展).
Secured two nationally funded projects as principal investigator, supported by NSFC and CAST.
DBConformer was accepted by IEEE Journal of Biomedical and Health Informatics.
MVCNet was accepted by Knowledge-Based Systems.
Received the Ph.D. National Scholarship (国家奖学金) for the second consecutive year.
Named a HUST Merit Student Pacesetter (华中科技大学三好学生标兵), the university’s highest student honor.
Named one of ten worldwide recipients of the inaugural Ant Group InTech Scholarship.
Canine EEG Helps Human was accepted by National Science Review.
CSDA was accepted by Knowledge-Based Systems.
Received the 2024 Ph.D. National Scholarship (国家奖学金).
Cross-species · cross-modality · cross-dataset · cross-patient
National Science Review
The first systematic demonstration that heterogeneous canine/human and intracranial/scalp EEG can improve cross-species seizure detection, achieving over 90% AUC with very limited target labels.
IEEE Transactions on Biomedical Engineering
A two-stage framework that combines cross-dataset teacher pretraining with cross-subject online adaptation through logit distillation and feature-level contrastive alignment.
Spatiotemporal modeling · contrastive learning · multimodal fusion
IEEE Journal of Biomedical and Health Informatics
Parallel temporal and spatial Conformer branches model long-range dynamics and channel interactions across three EEG paradigms, using over eight times fewer parameters than high-capacity Conformer baselines.
Science China Information Sciences
Band-specific encoders, frequency-aware attention, and auxiliary supervision improve auditory attention decoding across two datasets and three decision-window lengths.
ICLR 2027 submission
Reliability-aware EEG–EOG fusion combines band-aware Conformer and Mamba encoders, cross-modal attention, and sample-wise modality weighting.
IEEE Journal of Biomedical and Health Informatics
A knowledge-driven hybrid architecture combining a closed-form TDCA–TRCA front-end, learnable sub-band weighting, and an SSVEPformer decoder, with target-guided alignment for robust intra- and inter-subject decoding under limited calibration.
Data augmentation · generative models · trustworthy benchmarks
Artificial Intelligence Review
A unified taxonomy and benchmark spanning four BCI paradigms and four evaluation dimensions: signal realism, physiological plausibility, downstream utility, and privacy.
ICLR 2027 submission
CoMA-DiT treats paired modalities as mutual generative supervision through cross-modal attention and reliability-gated latent augmentation.
Trajectory
Young Student Basic Research Program (国家自然科学基金青年学生基础研究项目), Grant 625B2077, 2026–2027.
Young Elite Scientists Sponsorship Program, Doctoral Student Track (中国科协青年人才培育工程博士生专项), 2026–2027.
Algorithm Research Intern, Cloud-Server Failure Prediction
Open science
The open CHSZ dataset contains EEG recordings from 27 children with epilepsy and has received 2,182 downloads. Representative implementations are maintained openly on GitHub.
Recognition
Recognition for research originality, sustained academic excellence, and national and international distinction.
One of ten recipients worldwide in the inaugural cohort; featured by Zhejiang Daily (《浙江日报》).
Three consecutive awards during Ph.D. study (2024–2026); four-time recipient overall, including undergraduate study (2020).
Top 10 Advances in Brain–Computer Interfaces in China (2025), Top 10 Highlights (2024), and three national BCI competition awards: First Prize (2023), Second Prize (2021), and Second Prize (2025).
Awarded to the National Science Review paper Canine EEG Helps Human by the Hubei Association for Science and Technology (湖北省科协优秀科技论文奖).
Highest student honor at HUST.
One of five recipients worldwide.
Academic service
Reviewing across artificial intelligence, biomedical engineering, and neural engineering.
Reviewer for Artificial Intelligence Review (IF 18.8), IEEE Reviews in Biomedical Engineering (IF 16.0), IEEE Transactions on Fuzzy Systems, IEEE JBHI, IEEE TNSRE, IEEE Transactions on Biometrics, Behavior, and Identity Science, Knowledge-Based Systems, Neurocomputing, Journal of Neural Engineering, Scientific Data, ICONIP, and IEEE SMC. Recognized as an IOP Trusted Reviewer in 2024.
People
I have been fortunate to collaborate with researchers across transfer learning, neural decoding, multimodal learning, and biomedical AI.
Contact
I welcome conversations about cross-species learning, brain foundation models, brain-disorder intelligence, open datasets, and research internships.