PHYSTAT

# Sticking to the roots of machine learning: old-school approaches to physics analysis

## by Tommaso Dorigo (Universita e INFN, Padova (IT))

Europe/Zurich
Description

In this seminar I will take a step back from the recent trend of DNNizing everything, by discussing the merits of two recent results, produced using statistical learning approaches to supervised and unsupervised learning problems of interest to HEP. The first is the use of a 65-million-parameters k-NN algorithm for deep regression [1]. The second is the identification of anomalies in HEP data in a fully unsupervised way, exploiting the random sampling of their copula space [2].

Organized by

O. Behnke, L. Lyons, L. Moneta, N. Wardle

Videoconference
PHYSTAT
Zoom Meeting ID
68793225561
Host
Olaf Behnke
Passcode
07630691