Abstract:
This seminar explores the application of certifying algorithms to machine learning problems, namely linear separation.
After introducing a few preliminary notions, we present an efficient certifying algorithm that, given a set of n real points with binary labels, either returns a hyperplane separating the points or identifies d+2 of the labelled points that cannot be separated by any hyperplane using a constructive proof of Kirchberger's theorem.
Finally, a dimension-free and constructive extension of Kirchberger's theorem is also discussed and applied to find O(1/ε^2) of the labelled points that cannot be separated with a normalised margin ε by any hyperplane.
Il seminario è organizzato dai dottorandi di Matematica e si svolgerà in presenza presso il Dipartimento di Matematica e Fisica, via Della Vasca Navale 84, aula B.
Link identifier #identifier__150421-1Sito web
This post is also available in:
Link identifier #identifier__130194-5
Eng