A0429
Title: Robust probabilistic principal component analysis with mixture of exponential power distributions
Authors: Zhenghui Feng - Harbin Institute of Technology, Shenzhen (China) [presenting]
Xinyi Wang - Xiamen University (China)
Xiao Chen - HongKong Baptist University (Hong Kong)
Heng Peng - HongKong Baptist University (Hong Kong)
Abstract: The EP-MPPCA model is introduced, which serves as a flexible and robust alternative to conventional Gaussian-based mixtures of probabilistic principal component analysis (MPPCA) for high-dimensional data analysis. The EP-MPPCA model utilizes the exponential power distribution family, making it more adept at handling heterogeneous data distributions and outliers. Algorithms and estimation methods are provided for the EP-MPPCA model, and its performance is evaluated through simulations. In real data analysis, the EP-MPPCA model can be practically applied in two important applications: unsupervised clustering and image data reconstruction. Specifically, the EP-MPPCA model is shown to effectively handle outliers in high-dimensional image data, leading to improved reconstruction quality. Additionally, the model can achieve superior clustering results in an unsupervised manner for high-dimensional data.
