Neighborhood Component Analysis (Renaud Rivier, University of Geneva)

03.06.2019 17:00

Statistical learning is an interesting field in theoretical and applied mathematics; it combines ideas from geometry, probability theory and computer science (among others). In this talk, I would like to present a practical "data mining" challenge I worked on and some ideas that were very helpful in solving it - or at least in getting interesting results.

I will first briefly introduced statistical learning theory and present a few well-known algorithms. I will then move to a particular problem of fraud detection where classical algorithms fail to perform well. Finally I will talk about the idea of metric learning, present the method called "Neighborhood Component Analysis" and show how it could be used in the particular case of fraud detection.


Room 17, Séminaire des doctorants

Organisé par

Section de mathématiques


Renaud Rivier, University of Geneva

entrée libre


Catégorie: Séminaire

Mots clés: graduate seminar

Plus d'infos

Contact: missing email