BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Chao Tian (exas A&M University) DTSTART:20240313T160000Z DTEND:20240313T171500Z DTSTAMP:20260423T021444Z UID:AAIT/30 DESCRIPTION:Title: Co mputer-Aided Investigation of Information-Theoretic Limits: An Overview. ( 2nd part)\nby Chao Tian (exas A&M University) as part of Seminar on Al gorithmic Aspects of Information Theory\n\n\nAbstract\nThe linear programm ing (LP) formulation of information measures provides a solid mathematical framework to identify the fundamental limits of information systems compu tationally. A critical issue of this approach is however its high computat ional complexity. To reduce the computation burden of this approach\, we c an utilize the symmetry structure in such systems. The strength of the sym metry-reduced approach is illustrated in several well-known difficult prob lems\, such as regenerating codes\, coded caching\, and private informatio n retrieval\, which provides new and non-trivial outer bounds. In addition to rate bounds\, more in-depth studies can be conducted on the joint entr opy structure of these computed bounds\, which often lead to reverse-engin eered novel code constructions and further allow disproving linear code ac hievability. Finally\, we discuss two new directions: the first is to allo w the utilization of non-Shannon-type inequalities in the computational ap proach\, and the second is to convert the original LP into a sequence of s maller LPs\, both of which appear to be awaiting certain suitable machine- learning techniques.\n LOCATION:https://researchseminars.org/talk/AAIT/30/ END:VEVENT END:VCALENDAR