When random proportional subspaces are also random quotients (Q1883592)

From the authors' introduction: ``The present paper answers a vaguely put question from [\textit{J. Bourgain} and \textit{V. D. Milman}, Integral Equations Oper. Theory 9, 31--46 (1986; Zbl 0588.46011)], and opens a new direction in understanding what `random' subspaces and `random' quotients are. We consider a critical case and we show that if, for some specific random structure, a generic subspace of \textit{some} fixed proportional dimension is isomorphic (essentially the same) to a generic quotient of some proportional dimension of another space (with a similarly selected random structure) then for \textit{any} proportion arbitrarily close to \(1\), the first space has a lot of Euclidean subspaces and the second space has a lot of Euclidean quotients. So a complete similarity between a generic subspace and a generic quotient implies that most subspaces (respectively, quotients) are Euclidean.'' The method of proof is to use new properties of minimal and maximal volume ellipsoids (properties of considerable interest on their own) to show that the volume ratio (respectively, outer volume ratio) is bounded. This implies the existence of a large family of Euclidean subspaces (respectively, quotients).