Leonardo Cella

About me

I am a Postdoctoral Researcher @ Italian Institute of Technology where I have the great privilege to work with Massimiliano Pontil. Since November 2021 I joined the ERG-IIT jointlab investigating the benefit of machine-learning in the renewable evergies domain.
Before the postdoc I have been lucky to spend 3 great (and hard) years as a PhD Student @ University of Milan advised by Nicoló Cesa-Bianchi.

Together with Massimiliano Pontil we are investigating the multi-task and meta-learning settings with linear stochastic bandits. Recently, we had the pleasure to have Karim Lounici joining this research. We are now investigating the benefit of transfer-learning under a shared representation.

During summer 2018 I was selected for a four months internship at Amazon Berlin where I had the opportunity to work as Applied Scientist in the Music ML Team mentored by Giovanni Zappella.

Research Interests

I am interested in the theoretical aspects of machine learning solutions. So far, the area I investigated the more is multi-armed bandits. Along this direction, during my postdoc I have mainly investigated their combination with meta-learning approaches. I am very interested in the applications of machine-learning solutions and their usage in real-world scenarios.

Selected Publications

• L. Cella, K. Lounici and M. Pontil. Meta-learning Representation with Contextual Linear Bandits. Under review.
• L. Cella, K. Lounici and M. Pontil. Multi-task Representation Learning with Stochastic Linear Bandits. Under review.
• R. Grazzi, A. Akhavan, M. Pontil, I. J. Falk and L. Cella. Group Meritocratic Fairness in Linear Contextual Bandits. Under review.
• L. Cella and M. Pontil. Multi-Task and Meta-Learning with Sparse Linear Bandits.
  Proceedings of the 37th Conference on Uncertanity in Artificial Intelligence (UAI), 2021.
• L. Cella, C. Gentile and M. Pontil. Best Model Identification: a Rested Bandit Formulation.
  Proceedings of the 38th International Conference on Machine Leraning (ICML), 2021.
• L. Cella, A. Lazaric and M. Pontil. Meta-learning with Stochastic Linear Bandits.
  Proceedings of the 37th International Conference on Machine Learning (ICML), 2020
• L. Cella and N. Cesa-Bianchi. Stochastic Bandits with Delay-Dependent Payoffs.
  Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS), 2020
• I. Kuzborskij, L. Cella and N. Cesa-Bianchi. Efficient Linear Bandits through Matrix Sketching.
  Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics (AISTATS). PMLR, 89:177-185, 2019
• L. Cella, S. Cereda, M. Quadrana and P. Cremonesi. Deriving item features relevance from past user interactions.
  Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization (UMAP). Pages:275-279, 2017

News

• M.F.O May 2022, Oberwolfach. I attended the prestigious workshop Re-thinking high-dimensional mathematical statistics.
• C.I.R.M. December 2021, Marseille. I attended the Meeting in Mathematical Statistics.
• Neurips 2021 December 2021, virtual conference. I has been selected as outstanding reviewer.
• ICML 2021 July 2021, virtual conference. I presented the work entitled Best Model Identification: A Rested Bandit Formulation, which has been done in collaboration with Massi and Claudio.
• UAI 2021 July 2021, Virutal conference. I presented the work entitled Multi-Task and Meta-Learning with Sparse Linear Bandits, which has been done in collaboration with Massi.
• October 2020. I joined the Computational Statistics and Machine Learning lab @ Italian Institute of Technology in Genoa. There I will work on the intersection of meta-learning and stochastic bandits together with Massimiliano Pontil.
• ICML 2020 June 2020. I presented the work entitled Meta-learning with Stochastic Linear Bandits, which has been done in collaboration with Alessandro Lazaric and Massimiliano Pontil.
• AISTATS 2020 June 2020, Virtual conference. I presented the project inspired by the Amazon experience: Stochastic Bandits with Delay-Dependent Payoffs.
• 7th workshop on Stochastic Methods in Game Theory July 2020, Erice (Italy). I will give a poster presentation, I have not decided the topic yet.
• MAPLE Workshop, September 2019, Milan (Italy). I gave a talk on my research project Stochastic Bandits with Delay-Dependent Payoffs.
• AISTATS 2019 April 2019, Okinawa (Japan). I presented the research project done with Ilja and proposed by Nicoló.
• Applied Scientist Intern @ Amazon Berlin, July-October 2018.
• Reviewer for: AISTATS 2021, Neurips 2020, ALT 2019, Neurips 2019, AISTATS 2018.

Other Interests

I am strongly interested in the alleviation of human trafficking and I recently join cooperativa cielo to cooperate against that. I am also sensible to themes regarding barriers to education and learning in children. In Genoa I am a member of Sant'Egidio community, where I take care of these principles.