Dirichlet Process Gaussian Mixture

Dirichlet Process Gaussian Mixture Models (DPGMMs) are nonparametric Bayesian models used for clustering data where the number of clusters is unknown, offering a significant advantage over traditional methods requiring pre-specified cluster counts. Current research focuses on improving the efficiency and robustness of DPGMM inference, particularly addressing challenges like slow convergence and sensitivity to outliers through techniques such as variational inference, ensemble methods (e.g., random subspace and subsampling), and deep learning-based enhancements to sampling algorithms. These advancements are impacting diverse fields, including single-cell RNA sequencing analysis, outlier detection, and hyperspectral image segmentation, by enabling more accurate and automated clustering of complex datasets.

Papers

February 6, 2024

The VampPrior Mixture Model
Andrew Stirn, David A. Knowles
Variational Inference Latent Variable Model scRNA Seq Data Variational Mixture Dirichlet Process Gaussian Mixture

January 1, 2024

Unsupervised Outlier Detection using Random Subspace and Subsampling Ensembles of Dirichlet Process Mixtures
Dongwook Kim, Juyeon Park, Hee Cheol Chung, Seonghyun Jeong
Subsampling Method Unsupervised Outlier Detection Random Subspace Dirichlet Process Mixture Model Dirichlet Process Gaussian Mixture

May 23, 2023

DIVA: A Dirichlet Process Mixtures Based Incremental Deep Clustering Algorithm via Variational Auto-Encoder
Zhenshan Bing, Yuan Meng, Yuqi Yun, Hang Su, Xiaojie Su, Kai Huang, Alois Knoll
Generative Question Variational Auto Diva 360 Dirichlet Process Gaussian Mixture

August 12, 2022

EEGNN: Edge Enhanced Graph Neural Network with a Bayesian Nonparametric Graph Model
Yirui Liu, Xinghao Qiao, Liying Wang, Jessica Lam
Graph Neural Network Message Passing Message Passing Graph Neural Network Deep Graph Dirichlet Process Gaussian Mixture

March 25, 2022

Common Failure Modes of Subcluster-based Sampling in Dirichlet Process Gaussian Mixture Models -- and a Deep-learning Solution
Vlad Winter, Or Dinari, Oren Freifeld
Deep Learning Data Clustering Failure Mode Random Initialization Based Sampling Dirichlet Process Gaussian Mixture

March 5, 2022

Evaluation of Dirichlet Process Gaussian Mixtures for Segmentation on Noisy Hyperspectral Images
Kiran Mantripragada, Faisal Z. Qureshi
Global Evaluation Segmentation Based Approach Hyperspectral Image Image Segmentation Segmentation Result Segmentation Method Hyperspectral Image Denoising Dirichlet Process Gaussian Mixture

Dirichlet Process Gaussian Mixture

Papers

The VampPrior Mixture Model

Unsupervised Outlier Detection using Random Subspace and Subsampling Ensembles of Dirichlet Process Mixtures

DIVA: A Dirichlet Process Mixtures Based Incremental Deep Clustering Algorithm via Variational Auto-Encoder

EEGNN: Edge Enhanced Graph Neural Network with a Bayesian Nonparametric Graph Model

Common Failure Modes of Subcluster-based Sampling in Dirichlet Process Gaussian Mixture Models -- and a Deep-learning Solution

Evaluation of Dirichlet Process Gaussian Mixtures for Segmentation on Noisy Hyperspectral Images