Colloquium
Jeff M. Phillips
Duke University
Wednesday, April 15, 2009
3147 MEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Title: Handling Spatial Uncertainty in Data
Jeff M. Phillips
Duke University
Wednesday, April 15, 2009
3147 MEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Title: Handling Spatial Uncertainty in Data
Abstract
Many sensed data sets have inherent positional uncertainty. That is, the true location of each data point is not known precisely, but rather can be described by a probability distribution. For instance, the location of atoms in a protein structure, height values in a terrain, or obstacle locations sensed from a robot's laser range finder may be modeled with a probability distribution based on the error model from the sensing technology.
How should we compute statistics on these uncertain data sets? How should we visualize these data sets and the induced statistics? I will provide a simple and practical framework for handling these sorts of questions. The answers will themselves be distributions and will have approximation guarantees. I will focus on examples in geometric shape fitting, but the framework and its guarantees are more general.
Joint work with Maarten Loffler.
Many sensed data sets have inherent positional uncertainty. That is, the true location of each data point is not known precisely, but rather can be described by a probability distribution. For instance, the location of atoms in a protein structure, height values in a terrain, or obstacle locations sensed from a robot's laser range finder may be modeled with a probability distribution based on the error model from the sensing technology.
How should we compute statistics on these uncertain data sets? How should we visualize these data sets and the induced statistics? I will provide a simple and practical framework for handling these sorts of questions. The answers will themselves be distributions and will have approximation guarantees. I will focus on examples in geometric shape fitting, but the framework and its guarantees are more general.
Joint work with Maarten Loffler.
Bio
Jeff M. Phillips is currently a Postdoctoral Associate at Duke University. He received his Ph.D. in Computer Science from Duke in January, 2009 under the supervision of Pankaj Agarwal. Jeff's main interests are in computational geometric and statistical problems. Before his doctoral studies, Jeff received a BS in Computer Science and BA in Mathematics from Rice University in 2003. While at Rice, he worked with Lydia Kavraki on robotics.
Jeff M. Phillips is currently a Postdoctoral Associate at Duke University. He received his Ph.D. in Computer Science from Duke in January, 2009 under the supervision of Pankaj Agarwal. Jeff's main interests are in computational geometric and statistical problems. Before his doctoral studies, Jeff received a BS in Computer Science and BA in Mathematics from Rice University in 2003. While at Rice, he worked with Lydia Kavraki on robotics.
Return to
2009 Events Calendar