Title: Estimating the number of clusters for high-dimensional mixture model

Speaker:   Prof. Yiming Liu  (Nanyang Technological University)

Time: Thur, Apr.15 2021, AM: 9:30-10:10

Location: Tencent Conference

Meeting  Number: 749351635

Inviter: Dr. Shaochen Wang

Abstract:

      This talk proposes the variant Akaike and Bayesian information criteria to estimate the number of clusters for high dimensional mixture data. By investigating the limiting behaviors of the eigenvalues of sample covariance matrix for the observed data, the consistency of our criteria is obtained in different scenarios under some derived conditions. Simulation studies demonstrate that our approaches are more robust and competitive than other existing techniques.