Andrea Tirinzoni

Postdoctoral Researcher at INRIA Lille

I am a postdoctoral researcher in the SCOOL team (previously called SequeL) at INRIA Lille. I received my Ph.D. in computer science at Politecnico di Milano (Italy) advised by Marcello Restelli, where I defended the thesis "Exploiting Structure for Transfer in Reinforcement Learning". My main research interest is reinforcement learning.

My (probably not) updated CV can be downloaded here.


News

Postdoctoral position at INRIA Lille

Starting from April 2021, I am a postdoctoral researcher at INRIA Lille (SCOOL team).

Apr 2021
PhD thesis defense

I successfully defended my PhD thesis "Exploiting Structure for Transfer in Reinforcement Learning" and received a PhD cum laude in Information Technology.

Mar 2021
One paper accepted at Neurips 2020

Our paper An Asymptotically Optimal Primal-Dual Incremental Algorithm for Linear Contextual Bandits (with M. Pirotta, M. Restelli, and A. Lazaric) has been accepted for publication at Neurips 2020.

Sep 2020
One paper accepted at ICML 2020

Our paper Sequential Transfer in Reinforcement Learning with a Generative Model has been accepted for publication at the 37th International Conference on Machine Learning (ICML).

Jun 2020
Two papers accepted at AISTATS 2020

My colleagues and I got two papers accepted at AISTATS 2020: Truly Batch Model-Free Inverse Reinforcement Learning about Multiple Intentions (with G. Ramponi, A. Likmeta, A. Metelli, and M. Restelli) and A Novel Confidence-Based Algorithm for Structured Bandits (with A. Lazaric and M. Restelli).

Jan 2020
Internship at FAIR Paris

Starting from January until April 2020, I will be an intern at Facebook AI Research (Paris), supervised by Alessandro Lazaric. I will be working on exploration-exploitation in structured RL and bandit problems.

Dec 2019
One paper accepted at ICML 2019

Our paper Transfer of Samples in Policy Search via Multiple Importance Sampling has been accepted for publication at the 36th International Conference on Machine Learning (ICML).

May 2019
Teaching Assistant at RLSS 2019

I have been invited to serve as a teaching assistant at the Reinforcement Learning Summer SCOOL (RLSS). The RLSS is an in-depth course on RL and MABs organized by the SCOOL (formerly SequeL) research team from Inria. It will take place in Lille (France) from July 1st to July 12th.

April 2019
Two papers accepted at NeurIPS 2018

My colleagues and I got two papers accepted at NeurIPS 2018: Transfer of Value Functions via Variational Methods (with Rafael Rodriguez and Marcello Restelli) and Policy-Conditioned Uncertainty Sets for Robust Markov Decision Processes (with Xiangli Chen, Marek Petrik, and Brian Ziebart).

September 2018
Oral presentation at EWRL 2018

I will present my work Transferring Value Functions via Variational Methods at the 14th European Workshop on Reinforcement Learning on October 3rd in Lille.

September 2018

Publications

Preprints

A Fully Problem-Dependent Regret Lower Bound for Finite-Horizon MDPs

Andrea Tirinzoni, Matteo Pirotta, and Alessandro Lazaric

arXiv preprint

[pdf]

Conference Papers

Leveraging Good Representations in Linear Contextual Bandits

Matteo Papini, Andrea Tirinzoni, Marcello Restelli, Alessandro Lazaric, and Matteo Pirotta

International Conference on Machine Learning (ICML), 2021.

[pdf]

Meta-Reinforcement Learning by Tracking Task Non-stationarity

Riccardo Poiani, Andrea Tirinzoni, and Marcello Restelli

International Joint Conference on Artificial Intelligence (IJCAI), 2021.

[pdf]

An Asymptotically Optimal Primal-Dual Incremental Algorithm for Linear Contextual Bandits

Andrea Tirinzoni, Matteo Pirotta, Marcello Restelli, and Alessandro Lazaric

Advances in Neural Information Processing Systems 33 (NeurIPS) , Vancouver (Canada), 2020.

[pdf]

Sequential Transfer in Reinforcement Learning with a Generative Model

Andrea Tirinzoni, Riccardo Poiani, and Marcello Restelli

International Conference on Machine Learning (ICML), Vienna (Austria), 2020.

[pdf][slides] [code]

A Novel Confidence-Based Algorithm for Structured Bandits

Andrea Tirinzoni, Alessandro Lazaric, and Marcello Restelli

International Conference on Artificial Intelligence and Statistics (AISTATS), Palermo (Italy), 2020.

[pdf][slides]

Truly Batch Model-Free Inverse Reinforcement Learning about Multiple Intentions

Giorgia Ramponi, Amarildo Likmeta, Alberto Maria Metelli, Andrea Tirinzoni, and Marcello Restelli

International Conference on Artificial Intelligence and Statistics (AISTATS), Palermo (Italy), 2020.

[pdf]

Gradient-Aware Model-Based Policy-Search

Pierluca D'Oro, Alberto Maria Metelli, Andrea Tirinzoni, Matteo Papini, and Marcello Restelli

AAAI Conference on Artificial Intelligence, New York, 2020.

[pdf]

Transfer of Samples in Policy Search via Multiple Importance Sampling

Andrea Tirinzoni, Mattia Salvini, and Marcello Restelli

International Conference on Machine Learning (ICML), Long Beach, 2019.

[pdf] [poster] [slides] [code]

Feature Selection via Mutual Information: New Theoretical Insights

Mario Beraha, Alberto Maria Metelli, Matteo Papini, Andrea Tirinzoni, and Marcello Restelli

International Joint Conference on Neural Networks (IJCNN), Budapest, 2019.

[pdf]

Policy-Conditioned Uncertainty Sets for Robust Markov Decision Processes

Andrea Tirinzoni, Xiangli Chen, Marek Petrik, and Brian D. Ziebart

Advances in Neural Information Processing Systems (NeurIPS), Montreal, 2018. Spotlight (acceptance rate < 4%).

[pdf] [poster] [slides]

Transfer of Value Functions via Variational Methods

Andrea Tirinzoni, Rafael Rodriguez, and Marcello Restelli

Advances in Neural Information Processing Systems (NeurIPS), Montreal, 2018.

[pdf] [poster] [code]

Importance Weighted Transfer of Samples in Reinforcement Learning

Andrea Tirinzoni, Andrea Sessa, Matteo Pirotta, and Marcello Restelli

International Conference on Machine Learning (ICML), Stockholm, 2018.

[pdf] [poster] [slides] [code]

Journal Papers

Dealing with Multiple Experts and Non-stationarity in Inverse Reinforcement Learning: an Application to Real-life Problems

Amarildo Likmeta, Alberto Maria Metelli, Giorgia Ramponi, Andrea Tirinzoni, Matteo Giuliani, and Marcello Restelli

Machine Learning, 2021

[link]

Combining Reinforcement Learning with Rule-based Controllers for Transparent and General Decision-making in Autonomous Driving

Amarildo Likmeta, Alberto Maria Metelli, Andrea Tirinzoni, Riccardo Giol, Marcello Restelli, and Danilo Romano

Robotics and Autonomous Systems, 2020

[link]

Workshop Papers

Transferring Value Functions via Variational Methods

Andrea Tirinzoni, Rafael Rodriguez, and Marcello Restelli

European Workshop on Reinforcement Learning, Lille, 2018. Oral.

[link] [poster] [slides] [code]

Policy-Conditioned Uncertainty Sets for Robust Markov Decision Processes

Andrea Tirinzoni, Xiangli Chen, Marek Petrik, and Brian D. Ziebart

Workshop on Planning and Learning (PAL) @ ICML 2018. Oral.

[link] [poster] [slides]


Teaching

Informatica

Teaching Assistant

Bachelor's degree in Civil Engineering. Politecnico di Milano (Leonardo).

Course website on beep

a.y. 2019/2020, first semester
Informatica B

Laboratory Assistant

Bachelor's degree in Mechanical Engineering. Politecnico di Milano (Bovisa).

Course website

a.y. 2018/2019, first semester
Web and Internet Economics

Teaching Assistant

Master's degree in Computer Science and Engineering. Politecnico di Milano (Como).

Course website

a.y. 2017/2018, second semester
Informatica B

Laboratory Assistant

Bachelor's degree in Mechanical Engineering. Politecnico di Milano (Bovisa).

Course website

a.y. 2017/2018, first semester

Contacts

andrea [dot] tirinzoni [at] inria [dot] fr

Research Center of Inria Lille - Nord Europe

40, Avenue du Halley

59650 Villeneuve d’Asq, France