2008
********************************************************************************
Date :
August 22, Friday (3:30 pm)
Speaker : Lynne Seymour
(Department
of Statistics, University of Georgia)
Title : Probability Maps for Brain Activity via fMRI
Abstract :
We build a probability map of
the brain based on fMRI data from multiple subjects. The
model used is taken in part from particle interactions
models in physics and in part from social network models.
The neighborhood system for the social network is built
using a principal components breakdown of the image. The fit
is achieved creatively, using Markov chain Monte Carlo
methods as well as pseudo-likelihood methods. Preliminary
results are pleasing!
Place : PK 249
********************************************************************************
Date :
September
12, Friday (3:30 pm)
Speaker : Jeongyoun Ahn
(Department
of Statistics, University of Georgia)
http://aaron.stat.uga.edu/people/faculty/AHN/index.html
Title :
Geometry-Based Kernel Selection for Classification
Abstract :
Kernel-based classification
methods such as support vector machine use the data embedding idea
which maps data into a higher dimensional feature space via a kernel
function. In practice, selecting the hyperparameter in a kernel
function is a critical step in order to obtain a good performance.
We derive a novel method of choosing the hyperparameter in the
Gaussian kernel by considering the geometry of the embedded feature
space. The proposed method is independent of the specific
classification algorithm and empirically shown to give competitive
performance, while having much better computational efficiency than
cross-validation. We also demonstrate with data examples that the
proposed method is robust to the sampling variability, unlike
cross-validation.
Place : PK 249
********************************************************************************
Date :
October 31, Friday (3:30 pm)
Speaker : Paul Kvam (School
of Industrial Systems Engineering,
Georgia
Institute of
Technology)
http://www2.isye.gatech.edu/people/faculty/Paul_Kvam/
Title
:
Length Bias in the Measurements of
Carbon Nanotubes
Abstract :
To
measure carbon nanotube lengths, atomic force microscopy and special
software are used to identify and measure nanotubes on a square
grid. Current practice does not include nanotubes that cross the
grid, and, as a result, the sample is length-biased. The selection
bias model can be demonstrated through Buffon’s needle problem,
extended to general curves that more realistically represent the
shape of nanotubes observed on a grid. In this article, the
nonparametric maximum likelihood estimator is constructed for the
length distribution of the nanotubes, and the consequences of the
length bias are examined. Probability plots reveal that the
corrected length distribution estimate provides a better fit
to the Weibull distribution than the original selection-biased
observations, thus reinforcing a previous claim about the underlying
distribution of synthesized nanotube lengths.
Place : PK 249
********************************************************************************
Date :
November
14, Friday (3:30 pm)
Speaker : Xiangrong
Yin (Department
of Statistics, University of Georgia)
http://aaron.stat.uga.edu/~xryin/
Title
:
Longitudinal Data Analysis using Sufficient Dimension
Reduction
Method
Abstract :
There have been an
increasing number of applications where the data are
repeatedly measured at a sequence of time points. In this talk we
investigate a sufficient dimension reduction approach for analyzing
such longitudinal data. The proposed method can effectively reduce
the dimension of the predictors while retaining full regression mean
information. Variable selection for longitudinal data is studied,
and graphical diagnosis and model fitting after dimension reduction
are discussed. The proposed method is flexible enough to encompass a
variety of commonly used longitudinal models. Effectiveness of the
method is demonstrated through an extensive numerical study.
(Joint
work with Lexin Li at NCSU)
Place : PK 249
********************************************************************************
|
