Randomness and pattern in convex geometric analysis (Q1126746)

The author extrapolates in a philosophical manner the mathematical theory presented in this survey to the study of the World, in a final remark: ``I see the results of this theory as a ``window'' to the World of very high degree of freedom, just examples of organized behavior we should expect in the study of that World; not a chaotic diversity, exponentially increasing with increasing degree of freedom (=dimension in the presented theory), but on the contrary, an asymptotically well organized World with ``residual freedom'' reflected in our theory in a ``uniformly isomorphic'' view on the results.'' The survey begins with a list of titles about the local theory and the asymptotic theory of finite-dimensional normed spaces. As an example of very regular asymptotic behavior of an arbitrary high dimensional space is the study of convex bodies in sense to discover a pattern through the related properties. The first example is to construct some special ellipsoid \(\mathcal E\) which represents the body \(K\) or, to change the position of \(K\) by \(u \in SL_n\) such that \(u\mathcal E=\lambda D\), where \(D\) is the standard Euclidean ball and \(\lambda D\) represents \(uK\). For \(X=( \mathbb R^{n},\|\cdot\|,|\cdot|)\) where \(|\cdot|\) is the Euclidean norm, there are considered \[ k(X)= \max\left\{k\in \mathbb Z\mid\mu_{G_{n,k}} \left(\left\{E \in G_{n,k}\mid\tfrac{M}{2} | x| \leq \| x\| \leq 2M| x |, x \in E\right\}\right)> 1-\tfrac{k}{n-k}\right\} \] where \(\mu_{G_{n,k}}\) is the Haar probability measure on the Grassmannian manifold \(G_{n,k}\) of all k-dimensional subspaces of \(\mathbb R^n\), and \[ t(X)=\min\left \{t \in \mathbb Z\mid\exists u_i \in O(n)\text{ and }\tfrac{M}{2} | x| \leq \tfrac{1}{t}\sum_{i=1}^{t}\| {u_i(x)}\| \leq 2M| x | , x\in \mathbb R^n\right\}. \] \(k(X)\) is a ``local'' parameter because it describes the behavior of the subspaces of a space which belongs to a set of properties, and \(t(X)\) is a ``global'' parameter. The local and global parameters verify \(C_1n \leq k(X)t(X) \leq C_2n\), where \(C_1\), \(C_2\) are two universal constants. Further there are studied: the radius of the largest ball inscribed into the random projection \(P_EK\) (in the dual form), the geometric structure of the level sets \(K \bigcap rS^{n-1}\) that appear very similar for some values of \(r\), the bound number of the euclidean ball and for any other convex body K, the \(M\)-position of a convex body \(K\) giving the correct balance between K and the euclidean ball. In the 8-th section there are presented some recent developments in Local Theory and Convexity, respectively: Brascamp-Lieb inequalities and their applications, Economic embedding of n-dimensional subspaces of \(L_q\) to \(l_p^N\), Extension of the Dvoretzky-Rogers Lemma and corresponding factorization results. In the last section there are formulated some remarks about the isotropic positions in convex geometry. \(K\) is isotropic iff \(| K | =1\) and there is a constant \(L\) such that \(\int_{K}(f,x)(x,\varphi)dx=L(f,\varphi)\) for any \(f\) and \(\varphi\) in \(\mathbb R^n\). The \(M\)-position of \(K\) is an isotropic position.