Title:
A look at the applications of harmonic analysis on graphs to problems in signal processing
Abstract:
The basic ideas forming the core of harmonic analysis (the
relationships between smoothness, scale, and frequency) lift to the
setting of weighted graphs. we will explore how this can be used to solve
problems in audio and image denoising and machine learning.
Title:
Homogenization of Metric Hamilton-Jacobi Equations
Abstract:
The objective of this work is the effective numerical solution of front propagation problems in multiscale media.
We present a new approach which relates the cell problem for Hamitlon-Jacobi equation, the variational formulation for the
Lagrangian, and the variational formulation for the metric.
The main advantage of our approach is that we solve just one auxiliary equation to recover the homogenized Hamiltonian
$H(p)$. Previous methods require the solution of the cell problem (or a variational problem for each value of $p$.
Computational results are presented in the periodic case for the checkerboard pattern, and several other patterns. Exact solutions
are recovered numerically. We also present calculations in the random case.
This is joint work by Adam Oberman (Simon Fraser), Ryo Takei (UCLA), and Alex Vladimirsky (Cornell)
Title:
Fracture Paths from Front Kinetics: Relaxation and Rate-Independence
Abstract:
In many engineering models for fracture, crack fronts play a fundamental role: they are the
location of both crack growth and the energy dissipation due to growth.
However, there has not been a rigorous mathematical definition of crack
front, nor rigorous mathematical analysis predicting fracture paths using
these fronts as the location of growth and dissipation. I will discuss my
work on front based fracture models, which is joint with Chris Larsen (WPI) and
Michael Ortiz (Caltech). I will present a natural weak definition of crack front and
front speed, and discuss the analysis of models of crack growth in which the
energy dissipation is a (non-linear) function of the front speed.
Title:
Semidefinite programming methods for rank
minimization with applications in system theory
Abstract:
The nuclear norm (sum of singular values) of a matrix is often used
in convex heuristics for rank minimization problems in control, system
identification, signal processing, and statistics. These heuristics can
be seen as generalizations of 1-norm minimization methods for sparse
signal reconstruction.
In this talk we discuss the implementation of interior-point methods
for the solution of linear nuclear norm approximation problems.
This problem can be formulated as a semidefinite program that includes
large auxiliary matrix variables and is difficult to solve by general-purpose solvers. By exploiting problem structure, we show
that the cost per iteration of an interior-point method can be reduced
to roughly the cost of solving the approximation problem in Frobenius norm.
We will also discuss applications in system identification.
Title:
On Logical Types of Quantifiers in Natural Language
Abstract:
We review results on the logical types of quantifiers (Qs) expressible in English. Qs of
type (1,1) include all, some, most as in Most poets daydream. most poets itself is of
type (1). Complex (1,1) Qs include all but ten, more than ten, seven out of ten, vague
Qs (about a hundred), and non-permutation invariant Qs (John's two, as in John's two
children). We find one logical constraint that all such Qs satisfy: Conservativity. This
class (Cons) is provably the boolean closure of the generalized universal and
generalized existential Qs, themselves isomorphic to type (1) Qs. Cons includes
proportionality Qs, generally not first order definable (even over finite models), but
which do present novel entailment paradigms, ones that express the action of four
operations on Qs (identity, complement, post-complement and dual). We conclude with
some type (2) Qs, which express properties of binary relations, as in Different people
like different things or John danced with Mary but no one else danced with anyone else.
We prove a criterion for determining when type (2) Qs are Fregean reducible to
iterated application of type (1) Qs. Different-different and else-else are non-Fregean.
Title:
Hair Simulation at Walt Disney Animation Studios
Abstract:
Realistic hair modeling is a fundamental part of creating virtual
humans in computer graphics; however, realistically representing hair
in structure, motion and visual appearance is still an open challenge.
In the entertainment industry, such as with feature animation, it is
frequently a goal to model a physically impossible hairstyle or
motion. In these cases, a high degree of user control is needed to
direct the hair in a desired way, which can be a time-consuming and
costly endeavor due to the magnitude of the hair volume. In this talk,
I will present techniques used for accelerating hair simulation
through the use of level-of-detail algorithms as well as discuss the
practical day-to-day issues that arise in hair simulation for animated
films, specifically addressing hair simulation on Walt Disney
Animation Studios' upcoming film "Bolt".
Come to the graduate lounge to get tips and tricks from graduate students who have experience with the UCLA qualifying exam system. Learn about how much time to give yourself to prepare, what topics to focus on, good reference material, and more. Bring your questions and receieve your answers.
Panel discussion providing undergraduate students a chance to ask a panel of graduate students questions about their graduate school experience, the application process, funding, exams, and anything else they would like to know. The panel consists of a mix of both pure and applied math graduate students. This is a joint activity with the undergraduate student society.