Gexin Huang

Gexin Huang

Pattern Recognition, Bayesian Deep Learning, Generative Model, and Multi-Modal learning

About [Code] [CV] [Transcripts]

Currently, I am a PhD student supervised by Xiaoxiao Li at the Trusted and Efficient AI (TEA) lab, University of British Columbia.

During 2022-2023, I was a research assistant supervisored by Xiaodan Liang at the Human Cyber Physical Intelligence Integration (HCP) Lab, Sun Yat-Sen University (Guangzhou, China). Before that, I received master’s Degree in Pattern Recognition and Intelligent Systems at South China University of Technology (Guangzhou, China). Since 2018, I have been co-advised by Zhuliang Yu and Wei Wu at the Center for Brain-Computer Interfaces and Information Processing (director: Yuanqing Li).

My general research interests lie in Bayesian learning and machine learning techniques for interdisciplinary tasks. In particular, I am passionate about the intersection of deep learning and Bayesian statistics, geerative model, explainability, and graph representation learning.

Recent research works I focused on:

  • 2D&3D Brain Modelling via Diffusion Model [1]
  • Self-supervised learning and graph representation learning for whole slide images (WSIs) [2]
  • Ill-posed inverse problem for brain source localization via a deep learning paradigm[3]
  • EEG-fMRI multimodal learning with deep generative models and Bayesian inference [4]
  • EEG decoding and classification with hybrid models [5].

Publications

  • [1] Electromagnetic Source Imaging via a Data-Synthesis-Based Convolutional Encoder–Decoder Network
    Gexin Huang, Ke Liu, JW Liang, Zhu Liang Yu, Wei Wu, ZH Gu, FeiF Qi, and YQ Li.
    [IEEE Transactions on Neural Network and Learning System]
    [link] [code] [data]

  • [2] GeneFormer: Associating Biological Prior Relationships with Linguistic Knowledge for Explainable Pan-caner Genetic Multi-Label Classification
    Gexin Huang, ChenFei Wu, MingJie Li, Shen Zhao, Xiaojun Chang, Ying Sun, and XiaoDan Liang.
    [Under review on TPAMI]
    [link] [code] [data]

  • [3] Identity Preserving Diffusion Model for Brain Aging Modeling
    Gexin Huang, Mengwei Ren and Xiaoxiao Li.
    [OHBM 2024]
    [data]

Research Experiences

  • Winter 2023:
    • Topics: Brain Aging Modelling
    • Keywords: Diffusion model, Image-to-Image Generation, Video Generation and Interpolation.
    • Supervisor: Prof. Xiaoxiao Li
  • Winter 2022:
    • Topics: Explainable AI-driven Genomics Analysis for Computational Pathology
    • Keywords: graph representation learning, weakly and self-supervised learning, multi-label classification, explainability.
    • Supervisor: Prof. Xiaodan Liang
  • Fall 2021:
    • Topics: E/MEG Source Imaging via Multi-task Learning
    • Keywords: extreme multi-label learning, gated control network, graph embedding.
    • Supervisor: Prof. Wei Wu and Prof. Zhuliang Yu
  • Fall 2019:
    • Topic: Solving EEG Inverse Problem with Deep Learning Framework
    • Keywords: ill-posed, denoising autoencoder, data synthesis, knowledge-driven, spatio-temporal decomposition, Bayesian facoter model.
    • Supervisor: Prof. Wei Wu and Prof. Zhuliang Yu
  • Fall 2018:
    • Topic: Multi-modal Brain Signals Learning with Bayesian Deep Learning
    • Keywords: multi-view probabilistic model, deep generative model, variational inference, posterior regularization, nonparameteric eastimation.
    • Supervisor: Prof. Zhuliang Yu
  • Fall 2013: National College Students’ Innovation Program
    • Topic: Detection and Recognition of Ground Targets for Quadrotor UAV
    • Keywords: unmanned aerial vehicle, support vector machine, histogram of oriented gradients, data augmentation, objective detection and tracking.

Projects

  • [1] 2D&3D Brain Generation : Conditional Diffusion Model for Latent Space Manipulation [code]

  • [2] GeneFormer: Explainable Graph Representation Learning for Genomics Analysis of Biomedical Images [code]

  • [3] Data-synthesized Spatio-Temporal Convolutional Encoder–Decoder Network (DST-CedNet) [code]

  • [4] EEG-fMRI Multi-modal Learning [code]

  • [5] EEG Decodeing with Filter Bank Common Spatial Pattern Features [code]

  • [6] EEG Source Imaging with fMRI Priors [code]

Teaching and Academic Activities

  • Reviewer: TNNLS, CVPR 2022, CVPR 2023
  • Teaching assistant: Adavanced Techniques in Machine Learning, 2019 Fall
  • Teaching assistant: Digital Signal Processing, 2019 Winter

Skills

  • Matlab, C/C++, Java
  • Python
    • Numpy, Scipy, Pandas, HuggingFace
    • Keras, Tensorflow
    • Pytorch
  • Latex, Markdown, HTML, Ruby
  • Brainstorm, EEGLAB, SPM, OpenCV, OpenSlide