2009-01-29

generative model

Fergus and I talked for a while this morning about the Lang project of a generative model of the sky. Fergus opined that it would be an interesting project for computer vision, because it has a hope of being a complete model that generates all of the data, not just certain features or parts. On the other hand, it might become intractable fast, especially if we want the billions-by-billions covariance matrix. We discussed some of these issues.

No comments:

Post a Comment