Principles of Learning Problem Design: Call for Participation

NIPS 2007 Workshop

December 7-8, 2007 | Whistler, BC

Important Dates

Submission deadline: November 1, 2007

Notification: November 5, 2007

Workshop: December 7-8, 2007
Want to participate? If you have a relevant topic to discuss, email John Langford and Alina Beygelzimer. We'll be making decisions about the final program in early November.

Relevance is the most important aspect of any submission. Good relevant subjects include:

  1. Examples of learning problem design. What works in practice? Which problems are neglectable?
  2. Theory of learning problem design. How do we address these strange issues?
When we attempt to design learning problems, new issues arise that do not exist with conventionally labeled data. Examples of basic questions that we hope to address:

(1) How can we best deal with bias in our data source?

(2) How can we analyze the effectiveness of techniques for learning problem design?

(3) What are the core problems of learning problem design?

(4) What sort of modularity exists in learning problem design?

(5) What are good ancillary prediction problems? What core principles can be reused in designing learning problems?

(6) When problems are defined recursively, how do we insure that this definition leads to solutions the designer expects? How do errors propagate?