Fully lifted random duality theory
Abstract
We study a generic class of random optimization problems (rops) and their typical behavior. The foundational aspects of the random duality theory (RDT), associated with rops, were discussed in StojnicRegRndDlt10, where it was shown that one can often infer rops' behavior even without actually solving them. Moreover, StojnicRegRndDlt10 uncovered that various quantities relevant to rops (including, for example, their typical objective values) can be determined (in a large dimensional context) even completely analytically. The key observation was that the strong deterministic duality implies the, so-called, strong random duality and therefore the full exactness of the analytical RDT characterizations. Here, we attack precisely those scenarios where the strong deterministic duality is not necessarily present and connect them to the recent progress made in studying bilinearly indexed (bli) random processes in Stojnicnflgscompyx23,Stojnicsflgscompyx23. In particular, utilizing a fully lifted (fl) interpolating comparison mechanism introduced in Stojnicnflgscompyx23, we establish corresponding fully lifted RDT (fl RDT). We then rely on a stationarized fl interpolation realization introduced in Stojnicsflgscompyx23 to obtain complete statitionarized fl RDT (sfl RDT). A few well known problems are then discussed as illustrations of a wide range of practical applications implied by the generality of the considered rops.
Turn this paper into a full lesson
ArcXiv compiles a staged curriculum from this paper: 8-12 lessons across beginner → advanced, synthesised section guides, visuals, flashcards, a quiz, exercises, and on-demand deep dives per section. Grounded in the abstract, never invented.