Research

My research focuses on Statistical Reinforcement Learning both online and offline.

Left GIF

Google Deep Mind

Right GIF

Google Deep Mind

Papers

Articles in refereed journals

  1. Zangirolami, V., and Borrotti, M. (2024). Dealing with uncertainty: balancing exploration and exploitation in deep recurrent reinforcement learning. Knowledge-Based Systems, 293.

Book Chapters and proceedings

  1. Zangirolami, V., Pavesi, F. and Zanotti, M. (2026). Safe Exogenous State Reinforcement Learning for water tank system. Statistical Science: From Theory to Applied Research (SIS 2026). (accepted)

  2. Pedrazzini, E., Zangirolami, V., Migliorati, S. and Borrotti, M. (2026). Beta-Distributed Proximal Policy Optimization for Autonomous Driving with Grad-CAM Explainability. IFIP International Conference on Artificial Intelligence Applications and Innovations. (accepted)

  3. Pedrazzini, E., Zangirolami, V., Migliorati, S. and Borrotti, M. (2026). Flexible Beta: a Novel Policy Distribution for Proximal Policy Optimization. Statistical Science: From Theory to Applied Research (SIS 2026). (accepted)

  4. Zanotti, M., Zangirolami, V., and Pavesi, F. (2026). An evaluation of ensemble strategies for time series anomaly detection. Statistical Science: From Theory to Applied Research (SIS 2026). (accepted)

  5. Pavesi, F., Zanotti, M., and Zangirolami, V. (2026). A characterization of Gaussian processes over flat tori. Statistical Science: From Theory to Applied Research (SIS 2026). (accepted)

Conferences

Invited talk

  1. Zangirolami, V., Borrotti, M., and Candelieri, A.(2025). Conformal prediction for safe decision-making. HiTEc meeting and Workshop on Complex Data in Econometrics and Statistics, Limassol, Cyprus, 8th-9th July 2025 - Session chair

  2. Zangirolami, V., Borrotti, M., and Candelieri, A. (2025). Using Conformal Prediction for Gaussian Process Regression in Dynamic Systems. Statalk 2025, Milan, Italy, 13th-14th June 2025

  3. Zangirolami, V. (2024). Enhancing data efficiency in online deep reinforcement learning under partial observability. 18th International Conference of the ERCIM WG on Computational and Methodological Statistics, 13–16 December 2024, London (UK)