BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//MIDAS - ECPv5.1.5//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:MIDAS
X-ORIGINAL-URL:https://midas.umich.edu
X-WR-CALDESC:Events for MIDAS
BEGIN:VTIMEZONE
TZID:America/Detroit
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20170312T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20171105T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/Detroit:20170317T160000
DTEND;TZID=America/Detroit:20170317T170000
DTSTAMP:20211022T040718
CREATED:20160928T181331Z
LAST-MODIFIED:20190807T192929Z
UID:19065-1489766400-1489770000@midas.umich.edu
SUMMARY:Laura Balzano\, PhD\, University of Michigan - MIDAS Seminar Series
DESCRIPTION:\nLaura Balzano\, PhD\nAssistant Professor\nElectrical Engineering and Computer Science\nCollege of Engineering\nUniversity of Michigan\nSlides\n“Finding Low-Rank Structure in Messy Data”\n \nAbstract: In order to draw inferences from large\, high-dimensional datasets\, we often seek simple structure that model the phenomena represented in those data. Low-rank linear structure is one of the most flexible and efficient such models\, allowing efficient prediction\, inference\, and anomaly detection. In this talk we will cover at a high level some results in optimization and high-dimensional probability from the last several years that show how to identify low-rank structure in high-dimensional data despite missing and corrupted data. Additionally\, we will discuss two new directions for finding low-rank structure in messy real-world data. In the first\, we observe every entry of the matrix through a single unknown monotonic transformation. This is common in calibration and quantization problems. We show that matrix completion is still possible in this context and demonstrate a simple algorithm with guarantees. In the second\, our vector observations are heteroscedastic\, ie\, corrupted by one of several noise variances. This is common in problems like sensor networks or medical imaging\, where different measurements of the same phenomenon are taken with different quality sensing (eg high or low radiation). We prove recovery results for principal component analysis (PCA) in this context. We show that recovery for a fixed average noise variance is maximized when the noise variances are equal\, implying that while average noise variance is often a convenient measure of the overall quality of the data\, it gives an overly optimistic estimate of PCA performance. \nBio: Professor Balzano’s research projects are in statistical signal processing\, matrix factorization\, and optimization\, particularly dealing with large and messy data. She has worked in the areas of online algorithms\, non-convex formulations for matrix factorization\, compressed sensing and matrix completion\, network inference\, and sensor networks. Her interests are theoretical\, however\, her favorite mathematical problems are motivated by fascinating and important engineering problems. \n \nFor more information on MIDAS or the Seminar Series\, please contact midas-contact@umich.edu. MIDAS gratefully acknowledges Northrop Grumman Corporation for its generous support of the MIDAS Seminar Series. \n
URL:https://midas.umich.edu/event/laura-balzano/
LOCATION:Forum Hall\, Palmer Commons\, 100 Washtenaw Ave\, Ann Arbor\, MI\, 48109\, United States
CATEGORIES:MIDAS Weekly Seminar Series
END:VEVENT
END:VCALENDAR