MIT 6.854 Spring 2016 Lecture 19: Semidefinite Programming, MAXCUT

preview_player
Показать описание

Рекомендации по теме
Комментарии
Автор

4:50 - 13:30 MAXCUT problem (phrasing it as an (fractional) LP doesn't solve it more precisely than naive approximation algo with flipping a fair coin)
13:30 - 23:30 positive semidefinite (PSD) definition and SDP (semidefinite programming)
23:30 - 35:40 building intuition
35:40 - 41:20 lemma 1 (used for proving weak duality and another way of defining what a PSD is)
43:35 - 46:35 lemma 2 aka weak duality
47:15 - 49:50 strong duality "usually" holds (Slater's condition)
50:45 - 58:10 MAXCUT problem now solved using SDP
58:10 - 1:12:30 Goemans-Williamson theorem/analysis
1:12:30 UGC - unique game conjecture

aleksagordic
Автор

Students in this class are well prepared.

wuzhai