Erik Talvitie

I am currently a PhD student at the University of Michigan in the AI Lab. My advisor is Satinder Singh. As a group we are interested in the problem of creating broadly competant, adaptive agents. These days I am particularly interested in the idea of modular knowledge representation in which many simple models, each responsible for a restricted aspect of the world, are composed into a more complete model of the world that could be more compact, flexible and general than a flat representation could produce. In my current research I am exploring the idea of using models of abstractions of the world as these modules.

Publications

Maintaining Predictions Over Time Without a Model (pdf)
Erik Talvitie and Satinder Singh
Twenty-First International Joint Conference on Artificial Intelligence (IJCAI), 2009
(to appear)

Simple Local Models for Complex Dynamical Systens (pdf)
Erik Talvitie and Satinder Singh
Twenty-Second Annual Conference on Neural Information Processing Systems (NIPS), 2008

Building Incomplete but Accurate Models (pdf)
Erik Talvitie, Britton Wolfe, and Satinder Singh
Tenth International Symposium on Artificial Intelligence and Mathematics (ISAIM), 2008

An Experts Algorithm for Transfer Learning (pdf)
Erik Talvitie and Satinder Singh
International Joint Conferences on Artificial Intelligence (IJCAI), 2007
(pp. 1065-1070)