BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//NextToU//EN
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:ba770e08-1c7c-49b0-9fcd-9e019e769df8@nexttoyou
DTSTAMP:20261106T170752Z
DTSTART:20261117T220000Z
DTEND:20261117T230000Z
SUMMARY:CMDS - Reese Pathak
LOCATION:Blocker Building\, Blocker Building
DESCRIPTION:ABSTRACT\n\nMinimizing a loss over a set of data points is ubi
 quitous in data science. A complete understanding of which loss to minimi
 ze\, however\,\n\nis lacking. This talk focuses on an extremely simple mo
 del: estimating an unknown mean corrupted by Gaussian noise. An intuitive
 \n\nguess is that minimizing square loss (equivalently\, maximizing the l
 ikelihood) over the parameter space would be the best estimator\n\npossib
 le. Counterintuitively\, this is suboptimal in general. Embarrassingly\, 
 the reasons for why this occurs are not fully understood! I will\n\nexpla
 in some reasons why this can happen\, what can be done about it\, and som
 e related results around this theme. Based on several\n\npapers\, joint w
 ith Liviu Aoloritei (UC Berkeley)\, Cong Ma (U. Chicago)\, Michael I. Jor
 dan (UC Berkeley/INRIA)\, Annie Ulichney (UC\n\nBerkeley)\, and Nikita Zh
 ivotovskiy (UC Berkeley).\n\nhttp://next-to-u.app/events/ba770e08-1c7c-49
 b0-9fcd-9e019e769df8
URL:http://next-to-u.app/events/ba770e08-1c7c-49b0-9fcd-9e019e769df8
END:VEVENT
END:VCALENDAR