Abstracts
August 31
Speaker
Ron Peled
Title
On rough isometries of Poisson processes on the line
Abstract:
The notion of rough isometry (or quasi isometry) of metric spaces was
invented by Gromov in 81 and developed significantly by Kannai in 85.
It provides a way of saying that two metric spaces are somewhat
similar to each other, two spaces are rough isometric if their metric
is the same up to multiplicative and additive constants. For example,
R^2 and Z^2 are rough isometric.
The notion of rough isometry (or quasi isometry) of metric spaces was
invented by Gromov in 81 and developed significantly by Kannai in 85.
It provides a way of saying that two metric spaces are somewhat
similar to each other, two spaces are rough isometric if their metric
is the same up to multiplicative and additive constants. For example,
R^2 and Z^2 are rough isometric.
All concepts used will be defined, in particular no prior knowledge of
rough isometry is assumed
September 14
Speaker
Bradley Luen
Title
Ten things I wish I had known when I came to Berkeley
Abstract:
Berkeley can bewilder statisticians newly arrived from far-off
places. Many things I learned in later years, had I known them in those
crucial first months, could've prevented me from becoming the jaded,
burnt-out wreck I am today. In this talk I'll divulge some of these
crucial minutiae, from nonparametric testing to places to get good fruit
and vegetables. This short rant will be followed by a panel discussion,
during which new students can ask their aged and infirm seniors questions
about additional critical concepts. (This is conditional on me shaming
other old grads into being on the panel.)
September 14
Speaker
Daniel Ting
Title
Hierarchical Dirichlet Process and protein structure
Abstract:
Given the large existing body of work in DNA sequencing, research has
become increasingly focused on determining functions of proteins, the
products of genes. One such area of research is protein
structure prediction which aims to predict the 3D structure of a protein
based using computational methods. I will be describing one component used
in solving the protein structure
prediction problem and the methodology behind it.
The statistical problem is a density estimation problem for a class of
similar densities. Hierarchical Dirichlet Process (HDP) mixture models are
used to exploit similarities in the densities. I will give an overview of
HDPs and the related Chinese restaurant franchise process and present some
results from this model applied to the data. After the talk, we will be
going to Panda Express to simulate the Chinese restaurant franchise
process.