Alina and I are organizing a workshop on Learning Problem Design at NIPS.

**What is learning problem design?** It’s about being clever in creating learning problems from otherwise unlabeled data. Read the webpage above for examples.

**I want to participate!** Email us before Nov. 1 with a description of what you want to talk about.

Will the workshop touch on mechanism design for using human computation to build up labels? I’m sure everyone’s familiar with Luis von Ahn’s work in this area: http://www.cs.cmu.edu/~biglou/research.html

At any rate, this should be a great workshop…

I understand that, if human computation/labeling cost is quantified in “batch” sense (that is to ignore the labeling cognitive difficulty but only the number of labeled examples, label efficiency), “Learning Prob. Design” is the prob of Optimal Experiment Design/Active Learning. Am I wrong?

Just a minor administrative point: what’s the relationship between the Nov. 1 date and the Oct. 21 submission deadline on this page: http://hunch.net/~learning-problem-design/cfp.html ?

It’s sounding like a great workshop.

The relationship is “oops, I tweaked the webpage to be Nov. 1 now”.

I think of it a bit differently. Out there in the world is real problem (which we cannot design). This real problem must be mapped onto problems we know how to solve. There are several different ways to do this, and sometimes there are uncaptured issues. How to do that is what I think of as the essence.

We’re working on it.