Brown University Center for Statistical Sciences Seminar
Boston University School of Public Health, | |
Abstract: POSTPONED, POSTPONED Kaplan-Meier curves provide descriptors of survival information for different subgroups within a data set. Investigators frequently use these figures for descriptive comparison of the effect of a particular measure upon survival. When subjects enter a study at different ages and age is associated with survival, it is often desirable to adjust survival information for age differences in the subgroups. This talk will discuss several methods for age-adjustment of survival curves, including direct age-adjustment and proportional hazard modeling with age as a covariate. Examples will be presented from simulated data and from information collected in the Framingham Heart Study.
Stochastic Systems Seminar
Please note change in time and location for this week only! |
Abstract: Stability (or positive recurrence) of multiclass queueing networks has received considerable attention recently. Results on stability help to determine the maximum production rates of a manufacturing system that is modeled by a queueing network. These results are also essential in defining the notion of ``heavy traffic'' used in heavy traffic limit theorems. It is known that the stability of a queueing network is implied by the stability of a corresponding deterministic fluid limit model. Sharp stability regions for numerous queueing networks and their fluid models will be given. When the network parameters are in the stability region, a Lyapunov function is found to prove the stability of the fluid model, and hence that of the queueing network. When the network parameters are outside the stability region, the fluid network cycles to infinity, and the queueing network is transient.
Brown Applied Mathematics Pattern Theory and Vision Seminar
Abstract:
For signal and image classification and discrimination, in particular,
in medical or geophysical diagnostics and military applications,
extracting relevant features and reducing the dimensionality of measured
data is of vital importance.
As an attempt to automate the feature extraction procedure and to understand
what the critical features for these problems,
we developed the so-called local discriminant basis (LDB) method
which rapidly selects an orthonormal basis suitable for signal/image
classification problems from a large collection of orthonormal bases
(e.g., wavelet packets and local trigonometric bases).
Once the LDB is selected, a small number of most significant coordinates
(features) are fed into a traditional classifier such as linear discriminant
analysis or decision trees.
The performance of these statistical methods is enhanced since the
LDB method reduces the dimensionality of the problems without losing
important information for classification. Moreover, since the basis functions
well-localized in the time-frequency plane are used as feature extractors,
interpretation of the classification results becomes easier and more intuitive
than using the conventional methods.
In this talk, after briefly reviewing our toolkit (i.e., wavelet packets and local trigonometric functions), we describe the original LDB method (which maximizes certain "distances" among time-frequency energy distributions of signal classes) as well as its recent progress (which maximizes the "distances" among empirical probability densities of basis coordinates). We also show an application of these techniques to a real geophysical problem of classifying acoustic waveforms propagated in a borehole according to the lithology of geological formations. This talk will conclude with future directions of the adapted feature extraction technology, in particular, applications to cluster analysis of high dimensional datasets which turn out to have a deep connection with geometric analysis. (This is a joint work with Prof. Ronald R. Coifman at Yale University.)
P.D.E./Lefschetz Center Seminar
Department of Mathematics Colloquium
<--- 1997 Index