printable pdf
比利时vs摩洛哥足彩 ,
university of california san diego

****************************

math 288 - probability & statistics

garrett tresch

texas a&m university

stochastic embeddings of graphs into trees

abstract:

as the shortest path metric on a weighted tree can be embedded isometrically into a finite $\ell_1$ space, a lipschitz embedding of a given graph into $\ell_1$ can be obtained by constructing a low distortion embedding into a tree. conversely, while there are various topological properties of graphs that guarantee controlled distortion lipschitz embeddings into $\ell_1$ ($k$-outerplanar, series-parallel, low euler characteristic), it is still often the case that such a graph embeds quite poorly into a tree.

by introducing the notion of a stochastic embedding into a family of trees one can find more general concrete embeddings into $\ell_1$ then those limited by a single tree. in fact, it is known that every graph with n vertices embeds stochastically into trees with distortion o(log(n)). nevertheless, this upper bound is sharp for graphs such as expanders, grids and, by a recent joint work with schlumprecht, a large class of "fractal-like" series-parallel graphs called slash powers. 

in this talk we introduce an equivalent characterization of stochastic distortion called expected distortion and after proving a mild extension of a result of gupta regarding poor tree embeddings of a cycle, inductively lower bound the expected distortion of generalized laakso graphs found in most nontrivial slash power families.

host: chris gartland

may 16, 2024

11:00 am

apm 6402 https://ucsd.zoom.us/j/6806754343

****************************