|
Organizers |
Locating Boundaries of Machine Learning
by
Wesley Calvert
Southern Illinois University
Definitions for machine learnability are well-established. However, they can be difficult to check in a particular case. Much of the literature seems to consist of an ad-hoc algorithm for learning examples of a particular kind, proving that the respective class is learnable.
The main contribution of this talk is a precise calculation of the difficulty of determining whether a class is learnable or not.
On the other hand, the main technical challenge of that calculation is defining a topological setting sufficiently broad for the calculation to be meaningful, but sufficiently narrow for it to be possible.
Date received: December 2, 2016
Copyright © 2016 by the author(s). The author(s) of this work and the organizers of the conference have granted their consent to include this abstract in Topology Atlas. Document # cbnq-36.