BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Dissertation Defense: Rui Kang
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260909T220420Z
UID:tag:localist.com\,2008:EventInstance_47561562133461
DTSTART:20240920T160000Z
DTEND:20240920T180000Z
DESCRIPTION:Department of Statistics defense titled "Sparse Heteroskedastic
  PCA in High Dementions.\n\nAbstract:\n\nPrincipal component analysis (PCA
 ) is one of the most commonly used techniques for dimension reduction and 
 feature extraction. Though it has been well-studied for high-dimensional s
 parse PCA\, little is known when the noise is heteroskedastic\, which turn
 s out to be ubiquitous in many scenarios. We propose an iterative algorith
 m\, called SparseHPCA\, for the sparse PCA problem in the presence of hete
 roskedastic noise\, which alternatively updates the estimates of the spars
 e eigenvectors using orthogonal iteration with adaptive thresholdings in o
 ne step\, and imputes the diagonal values of the sample covariance matrix 
 to reduce the estimation bias due to heteroskedastic noise in the other st
 ep. Our procedure is computationally fast and provably optimal under the g
 eneralized spiked covariance model\, assuming the leading eigenvectors are
  sparse. A comprehensive simulation study shows its robustness and effecti
 veness under various settings. The application of our new method to two hi
 gh-dimensional genomics datasets\, i.e.\, microarray and single-cell RNA s
 equencing (scRNA-seq) data\, demonstrates its ability to preserve inherent
  cluster structures in downstream analyses. Additionally\, we apply Sparse
 HPCA to address the sparse singular value decomposition (sparse SVD) probl
 em in the presence of heteroskedastic noise\, further showcasing its versa
 tility.\n\nAdvisor: Dr. Zhao Ren
GEO:40.441386;-79.954582
LOCATION:Wesley W. Posvar Hall\, Statistics Seminar Room
SUMMARY:Dissertation Defense: Rui Kang
URL;VALUE=URI:https://calendar.pitt.edu/event/dissertation-defense-rui-kang
CATEGORIES:Defenses
END:VEVENT
END:VCALENDAR