A5026624008- Tensor decomposition and applications
- Complex Network Analysis Techniques
- Advanced Neuroimaging Techniques and Applications
- Advanced Clustering Algorithms Research
- Bayesian Methods and Mixture Models
- Face and Expression Recognition
- Computational Physics and Python Applications
Statistical Service
2025
Columbia University
2025
Hong Kong University of Science and Technology
2021-2023
University of Hong Kong
2021-2023
We study the problem of community detection in multilayer networks, where pairs nodes can be related multiple modalities. introduce a general framework, that is, mixture stochastic block model (MMSBM), which includes many earlier models as special cases. propose tensor-based algorithm (TWIST) to reveal both global/local memberships nodes, and layers. show TWIST procedure accurately detect communities with small misclassification error number and/or layers increases. Numerical studies confirm...
The latent class model is a widely used mixture for multivariate discrete data. Besides the existence of qualitatively heterogeneous classes, real data often exhibit additional quantitative heterogeneity nested within each class. modern analysis also faces extra challenges, including high-dimensionality, sparsity, and heteroskedastic noise inherent in Motivated by these phenomena, we introduce Degree-heterogeneous Latent Class Model propose an easy-to-implement HeteroClustering algorithm it....
We introduce a unified framework, formulated as general latent space models, to study complex higher-order network interactions among multiple entities. Our framework covers several popular models in recent analysis literature, including mixture multi-layer model and hypergraph model. formulate the relationship between positions observed data via generalized multilinear kernel link function. While our enjoys decent generality, its maximum likelihood parameter estimation is also convenient...
In this paper, we first study the fundamental limit of clustering networks when a multi-layer network is present. Under mixture stochastic block model (MMSBM), show that minimax optimal error rate, which takes an exponential form and characterized by Renyi divergence between edge probability distributions component networks. We propose novel two-stage method including tensor-based initialization algorithm involving both node sample splitting refinement procedure likelihood-based Lloyd...
We introduce a unified framework, formulated as general latent space models, to study complex higher-order network interactions among multiple entities. Our framework covers several popular models in recent analysis literature, including mixture multi-layer model and hypergraph model. formulate the relationship between positions observed data via generalized multilinear kernel link function. While our enjoys decent generality, its maximum likelihood parameter estimation is also convenient...