Friday, December 21, 2007

Interactive machine learning talk

Dan Olsen at BYU gave a talk, "Interactive Machine Learning", at UW CS a couple months back.

Dan's group is doing some clever work that combines machine learning and HCI. The UW CS talk is good but long. If you are short on time, first take a look at the fun short demo videos Dan's group produced.

I particularly recommend seeing the clever "Screen Crayons" (WMV) application for annotating documents and the "Teaching Robots to Drive" (WMV) demo of an intuitive interface for training a robot car. The "Image Processing with Crayons" (WMV) demo is also good for getting quick introduction to the core idea.


Amit said...

Kids often learn by imitation first, and then they generalize to learn a model or concepts. With machine learning it seems that people have focused on the model building part, because we tend to think of models as “intelligence” whereas imitation is not. But learning by example, and overgeneralizing, and then having the parent (teacher) correct that, seems like it matches the way kids learn, and I think they might learn faster this way than by learning the models directly. It's nice to see these videos; they not only demonstrate the concept, but the videos themselves are teaching us by example.

Julien Couvreur said...

The Tablet PC interface is an interesting example of some of the same principles discussed in this presentation.
The handwriting recognizer and the interface around it in Tablet SP1 have some feedback and learning built-in.
Yes, it is possible to personalize the recognizer up-front, by training it for 20 minutes. But most likely I don't want to spend that much time, and the UI actually supports that and learns incrementally by allowing errors in recognition to be corrected.

Julien Couvreur said...

I double-checked and the main improvements were actually in Tablet PC 2005 SP2, not SP1. Vista also has a bunch of improvements, if I recall.