Publications

[14] A Diffusion Approximation for Temporal-Difference Learning with Linear Features under Markovian Noise,
M. Forzo, A. Russo, E. Monzio Compagnoni, A. Pacchiano,
In Proc. of Decision-Making from Offline Datasets to Online Adaptation: Black-Box Optimization to Reinforcement Learning Workshop at ICML, 2026.
Download paper here

[13] Frontier Learning: Training LLM Reasoners at the Edge of Capability,
S. S. Ramesh*, R. Faro*, E. Monzio Compagnoni, I. Bogunovic, A. Lucchi,
In Proc. of Foundations of Deep Generative Models (FoGen) Workshop at ICML, 2026.
Download paper here

[12] Beyond a Single Explanation of the Adam–SGD Gap,
C. Zhang, R. Islamov, E. Monzio Compagnoni, J. Pang, A. Lucchi, A. Orvieto,
arXiv preprint arXiv:2606.14259, 2026.
Download paper here

[11] On the Interaction of Batch Noise, Adaptivity, and Compression, under $(L_0,L_1)$-Smoothness: An SDE Approach,
E. Monzio Compagnoni, R. Islamov, F. N. Proske, A. Lucchi, A. Orvieto, E. Gorbunov,
In Proc. of the Forty-Third International Conference on Machine Learning (ICML), 2026.
Download paper here

[10] Adaptive Methods Are Preferable in High Privacy Settings: An SDE Perspective,
E. Monzio Compagnoni, A. Stanghellini, R. Islamov, A. Lucchi, A. Koloskova,
In Proc. of International Conference on Learning Representations (ICLR), 2026.
Download paper here

[9] Unified Perspectives on Balancedness and Parameter-norm Evolution in Neural Nets,
J. Singh, E. Monzio Compagnoni, A. Orvieto,
In Proc. of Scientific Methods for Understanding Deep Learning Workshop at ICLR, 2026.
Download paper here

[8] On the Interaction of Noise, Compression Role, and Adaptivity under (L_0, L_1)-Smoothness: An SDE-based Approach,
E. Monzio Compagnoni, R. Islamov, A. Orvieto, E. Gorbunov,
In Proc. of High-dimensional Learning Dynamics at ICML, 2025.
Download paper here

[7] Unbiased and Sign Compression in Distributed Learning: Comparing Noise Resilience via SDEs,
E. Monzio Compagnoni, R. Islamov, F. N. Proske, A. Lucchi,
In Proc. of International Conference on Artificial Intelligence and Statistics (AISTATS), 2025 (Oral).
Download paper here

[6] Adaptive Methods through the Lens of SDEs: Theoretical Insights on the Role of Noise,
E. Monzio Compagnoni, T. Liu, R. Islamov, F. N. Proske, A. Orvieto, A. Lucchi,
In Proc. of International Conference on Learning Representations (ICLR), 2025.
Download paper here

[5] SDEs for Minimax Optimization,
E. Monzio Compagnoni, A. Orvieto, H. Kersting, F. N. Proske, A. Lucchi,
In Proc. of International Conference on Artificial Intelligence and Statistics (AISTATS), 2024.
Download paper here

[4] Risk Sharing with Deep Neural Networks,
M. Burzoni*, A. Doldi*, E. Monzio Compagnoni*,
In Journal of Quantitative Finance, 2024.
Download paper here

[3] An SDE for Modeling SAM: Theory and Insights,
E. Monzio Compagnoni, L. Biggio, A. Orvieto, F. N. Proske, H. Kersting, A. Lucchi,
In Proc. of 40th International Conference on Machine Learning (ICML), 2023.
Download paper here

[2] On the Effectiveness of Randomized Signatures as Reservoir for Learning Rough Dynamics,
E. Monzio Compagnoni, A. Scampicchio, L. Biggio, A. Orvieto, T. Hofmann, J. Teichmann,
In Proc. of International Joint Conference on Neural Networks (IJCNN), 2023.
Download paper here

[1] Empirics on the Expressiveness of Randomized Signature
E. Monzio Compagnoni, L. Biggio, A. Orvieto,
The Symbiosis of Deep Learning and Differential Equations: DLDE Workshop - NeurIPS, 2021.
Download paper here

* denotes equal contributions.