Obtaining fast error rates in nonconvex situations

  • Authors:

  • Shahar Mendelson

  • Affiliations:

  • Centre for Mathematics and its Applications, Institute of Advanced Studies, The Australian National University, Canberra, ACT 0200, Australia and Department of Mathematics, Technion, I.I.T, Haifa ...

  • Venue:
  • Journal of Complexity
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

We show that under mild assumptions on the learning problem, one can obtain a fast error rate for every reasonable fixed target function even if the base class is not convex. To that end, we show that in such cases the excess loss class satisfies a Bernstein type condition.