• View
• Edit
• History
• Print

Problems In Computational Ecology And Learning With Aggregate Data

Abstract: The goal of my research is develop algorithms to understand and make decisions about the environment using large data sets. In this talk, I will first summarize the different problems I work on, which include continent-scale modeling of bird migration, analysis of biological patterns in weather radar data, and network optimization models for conservation of species. Then, I will discuss some technical aspects of the problem of modeling bird migration, and how it has led to the invention of a new formalism called Collective Graphical Models for efficient probabilistic reasoning about large populations when only aggregate data is available. I will describe key ideas behind inference algorithms for Collective Graphical Models, highlight connections to other CS research problems, and discuss future challenges.