Andrea Locatelli, PhD student
at the University of Magdeburg.

I am was a PhD student at the Institut für Mathematische Stochastik in the mathematics deparment at the Otto-von-Guericke-Universität Magdeburg. I was previously in the same position at the University of Potsdam, hence my living in Berlin. My advisor is Alexandra Carpentier, who is the recipient of an Emmy Noether research grant. My interests are at the interface of multi-armed bandit, nonparametric statistics, active learning and optimization. I also keep an open eye on reinforcement learning and dynamic pricing.

Publications

• Rotting bandits are no harder than stochastic ones, AiStats 2019
  with Julien Seznec, Michal Valko, Alessandro Lazaric and Alexandra Carpentier.
  

• Active multiple matrix completion with adaptive confidence sets, AiStats 2019
  with Michal Valko and Alexandra Carpentier.
  

• Adaptivity to Smoothness in X-armed Bandits, COLT 2018
  with Alexandra Carpentier.
  

• An Adaptive Strategy for Active Learning with Smooth Decision Boundary, ALT 2018
  with Alexandra Carpentier and Samory Kpotufe.
  

• Adaptivity to Noise Parameters in Nonparametric Active Learning, COLT 2017
  with Alexandra Carpentier and Samory Kpotufe.
  

• Tight (lower) Bounds for the Fixed Budget Best Arm Identification Bandit Problem, COLT 2016
  with Alexandra Carpentier.
  

• An Optimal Algorithm for the Thresholding Bandit Problem, ICML 2016
  with Maurilio Gutzeit and Alexandra Carpentier.

Experience

• Visiting Student at Princeton University (October 2017 - December 2017): working on active learning and contextual bandits with Samory Kpotufe.

• Research Internship at University of Potsdam (October 2015 - May 2016): worked on my master's thesis for the MVA master at ENS Cachan under the supervision of Alexandra Carpentier and Michal Valko.
• Data Scientist at PriceMatch (June 2013 - February 2015): worked on dynamic pricing with a reinforcement learning and machine learning flavor for a french start-up that was recently acquired by Booking .