Lefschetz Center for Dynamical Systems Seminar
Brown University Center for Statistical Sciences Seminar
Abstract: In many epidemiologic studies the first indication of an environmental or genetic contribution to the disease is the way in which the diseased cases cluster within the same family units. We assume that all individuals are exchangeable, except for their disease status. This assumption of exchangeability is used to test the initial hypothesis of no familial link with the disease. Conditional on the distribution of the sizes of the various familial units, we obtain the exact probability of observing a given set of disease cluster frequencies. The expected frequencies and other moments are given. Two numerical examples demonstrate these methods. We describe an algorithm for obtaining exact statistical inference by enumerating all possible outcomes consistent with the numbers and sizes of the family units.
Stochastic Systems Seminar
Abstract:
We discuss how blocking phenomena in manufacturing, computer and communication networks can be modelled by networks of queues in
order to evaluate performance. Furthermore, new closed form solutions
to evaluate the throughput and normalization constants of certain
Markovian type queuing networks are
presented. We examine the case when detailed as well as when partial
balance conditions hold. LEMS and Electrical Science Seminar
LEMS and Electrical Science Seminar
Abstract:
The TrafficVision system is an advanced, next-generation video detector for
monitoring traffic on freeways and arterials. TrafficVision interprets live
roadway video to generate a host of traffic measurements such as vehicle
counts, speeds, lane changes, etc. for traffic control, road use monitoring
and traffic incident detection. The system makes use of a combination of
image processing techniques and the Ni1000 neural network chip to implement
real-time vehicle detection, tracking and classification. The system
executes on the Windows95 platform. This talk will present an overview of
the TrafficVision development effort, from initial project requirements to
an assessment of current performance. The talk will review the technical
approaches used in TrafficVision for camera calibration, vehicle detection,
tracking and classification. A demonstration of the system will be provided.
Brown Applied Mathematics Pattern Theory and Vision Seminar
Abstract:
In many estimation and recognition problems one needs to
incorporate global information in an efficient way.
Often this can be done by using a hierarchical
multistage approach where the initial stages are based on local
information and global constraints are introduced gradually as
one moves up in the hierarchy. Then, the estimation
process derives a hierarchy of distributed representations
containing local descriptors at the lowest levels and more
global ones at the higher levels.
In the first part of the talk, I will describe a general framework
to formulate distributed and hierarchical estimation problems.
Complex objects with large support are reconstructed from
localized descriptors by means of ``composition''.
The goal of estimation is to compute an efficient covering of the viable
configuration set, that is, an approximation of
every possible object which is consistent with the data.
Probabilistic modeling can also be included in this (mostly set-based)
method to represent uncertainty.
The second part of the talk will describe how these ideas can
be applied to edge detection. The estimation stage which derives
regular curves from tangent vectors will be described in some detail.
PDE/Lefschetz Center Seminar
Department of Mathematics Colloquium
<---
1997 Index
Real Time Traffic Monitoring