Seminar on Computational Learning and Adaptation


  Inducing Process Models from Continuous Data

Pat Langley
Institute for the Study of Learning and Expertise, and
Center for the Study of Language and Information, Stanford University
langley@isle.org
http://newatlantis.isle.org/~langley

In this talk, I pose a novel research problem for machine learning that involves constructing a quantitative process model from numeric data. I claim that casting learned knowledge in terms of processes with associated equations is desirable for scientific and engineering domains, where such notations are commonly used. I also argue that existing induction methods are not well suited to this task, although some techniques hold partial solutions. In response, I describe an approach to learning process models from time-series data and then illustrate its behavior in a population dynamics domain. In closing, I describe open issues in process model induction and encourage other researchers to tackle this important problem.

This talk describes work done jointly with Javier Sanchez, Ljupco Todorovski, and Saso Dzeroski. Details are available in the paper

http://www.isle.org/~langley/papers/process.icml02.ps

which will appear in the Proceedings of the Nineteenth International Conference on Machine Learning.



Date: Thursday, May 30

Time: 4:15-5:30PM

Place: Ventura 17


Return to the seminar schedule