Speaker: Zeyu
Guo

\nAffiliation: Ohio State University

Title: TBD

\nAbstract: TBD

\n END:VEVENT BEGIN:VEVENT UID:ai1ec-390@www.cs.jhu.edu/~mdinitz/theory DTSTAMP:20241012T214924Z CATEGORIES: CONTACT: DESCRIPTION:Speaker: Yuzhou Gu\nAffiliation:NYU Center for Data Science & C ourant Institute\nTitle: Community detection in the hypergraph stochastic block model\nAbstract:\nCommunity detection is a fundamental problem in ne twork\nscience\, and its theoretical study has received significant attent ion\nover the last decade. In this talk I will present some recent advance s\non the community detection problem in sparse hypergraphs. In\nparticula r\, we determine the weak recovery threshold for the\nhypergraph stochasti c block model for a wide range of parameters. This\nresolves conjectures m ade by physicists in the corresponding regimes\nand has implications to ph ase transitions of random constraint\nsatisfaction problems. A key compone nt in this study is to analyze the\nbehavior of information channels under repeated applications of the\nbelief propagation operator. We introduce a framework for performing\nthis analysis based on information-theoretical methods for channel\ncomparison. Along the way\, we formulate a rigorous v ersion of the\npopulation dynamics algorithm\, an approach commonly used i n practice\nbut lacks theoretical guarantees. DTSTART;TZID=America/New_York:20240925T120000 DTEND;TZID=America/New_York:20240925T130000 SEQUENCE:0 SUMMARY:[Theory Seminar] Yuzhou Gu URL:https://www.cs.jhu.edu/~mdinitz/theory/event/theory-seminar-yuzhou-gu/ X-COST-TYPE:free X-ALT-DESC;FMTTYPE=text/html:\\n\\n\\nSpeaker: Yuzh ou Gu

\nAffiliation:NYU Center for Data Science & Courant Institute< /p>\n

Title: Community detection in the hypergraph stochastic block mode l

\nAbstract:

\nCommunity detection is a fundamental problem i
n network

\nscience\, and its theoretical study has received signific
ant attention

\nover the last decade. In this talk I will present som
e recent advances

\non the community detection problem in sparse hype
rgraphs. In

\nparticular\, we determine the weak recovery threshold f
or the

\nhypergraph stochastic block model for a wide range of parame
ters. This

\nresolves conjectures made by physicists in the correspon
ding regimes

\nand has implications to phase transitions of random co
nstraint

\nsatisfaction problems. A key component in this study is to
analyze the

\nbehavior of information channels under repeated applic
ations of the

\nbelief propagation operator. We introduce a framework
for performing

\nthis analysis based on information-theoretical meth
ods for channel

\ncomparison. Along the way\, we formulate a rigorous
version of the

\npopulation dynamics algorithm\, an approach commonl
y used in practice

\nbut lacks theoretical guarantees.