Learning theory and generalization error bounds: Can we keep it simple?

Room A008

This talk will introduce basic concepts of learning theory in a tutorial-like manner. In particular, we will first see how to derive a simple generalization error bound with (almost) no background knowledge. Then, we will emphasize its limitations and sketch the ideas commonly used in statistical learning theory to overcome them, before concluding on whether it is possible to obtain practically relevant error bounds in a simple manner.

Slides