About Me

I am a Post Doc in the CSML group in IIT in Genoa and was previously a PhD Student at IIT and UCL under the supervision of Massimiliano Pontil.

I like to study and design efficient and scalable machine learning methods. The main focus of my research has been on efficient gradient-based bilevel optimization, with applications to hyperparameter optimization and meta-learning. In particular, I studied the rate of convergence in deterministic, stochastic and non-smooth settings.

More recently, I have been interested in in-context learning, in adaptive, parameter-free optimization techniques and in learning methods for dynamical systems.

News

[May 2024] Our paper “Nonsmooth Implicit Differentiation: Deterministic and Stochastic Convergence Rates” has been accepted at ICML 2024.
[May 2024] Our paper “Is Mamba Capable of In-Context Learning?” has been accepted at the conference on AutoML.
[April 2024] Julien and I gave a talk at the AutoML Online Seminar on the work “Is Mamba Capable of In-Context Learning?” (YouTube link) (Extended Slides)
[March 2024] Our paper “Is Mamba Capable of In-Context Learning?” has been accepted at the ICLR 2024 Workshop on Mathematical and Empirical Understanding of Foundation Models (ME-FoMo).

Seminars and Lectures

[April 2024] Is Mamba Capable of In-Context Learning?, AutoML Online Seminar (YouTube link) (Extended Slides)
[August 2023] Principled and Efficient Bilevel Optimization for Machine Learning, Lecture at the Bilevel Summer School (Slides) (School Program)

Publications

ICML

Nonsmooth Implicit Differentiation, Deterministic and Stochastic Convergence Rates

Riccardo Grazzi, Massimiliano Pontil, Saverio Salzo

Intenrnational Conference of Machine Learning, 2024.

PDF Code

AutoML

Is Mamba Capable of In-Context Learning?

Riccardo Grazzi*, Julien Siems*, Simon Schrodi, Thomas Brox, Frank Hutter (*equal contribution)

International Conference on Automated Machine Learning, 2024.

PDF Code

ICLR

Learning invariant representations of time-homogeneous stochastic dynamical systems

Vladimir R. Kostic, Pietro Novelli, Riccardo Grazzi, Karim Lounici, Massimiliano Pontil

International Conference on Learning Representations, 2024.

PDF Code Project Page

JMLR

Bilevel Optimization with a Lower-level Contraction: Optimal Sample Complexity without Warm-start

Riccardo Grazzi, Massimiliano Pontil, Saverio Salzo

Journal of Machine Learning Research, 2023.

PDF Code

NeurIPS

Group Meritocratic Fairness in Linear Contextual Bandits

Riccardo Grazzi, Arya Akhavan, John Isak Texas Falk, Leonardo Cella, Massimiliano Pontil

Neural Information Processing Systems, 2022.

PDF Code

Arxiv

Meta-Forecasting by combining Global Deep Representations with Local Adaptation

Riccardo Grazzi, Valentin Flunkert, David Salinas, Tim Januschowski, Matthias Seeger, Cedric Archambeau

PDF

AISTATS

Convergence Properties of Stochastic Hypergradients

Riccardo Grazzi, Massimiliano Pontil, Saverio Salzo

International Conference on Artificial Intelligence and Statistics, 2021.

PDF

ICML

On the Iteration Complexity of Hypergradient Computation

Riccardo Grazzi, Luca Franceschi, Massimiliano Pontil, Saverio Salzo

Intenrnational Conference of Machine Learning, 2020.

PDF Code

ICML

Learning-to-Learn Stochastic Gradient Descent with Biased Regularization

Giulia Denevi, Carlo Ciliberto, Riccardo Grazzi, Massimiliano Pontil

Intenrnational Conference of Machine Learning, 2019.

PDF Code

ICML

Bilevel Programming for Hyperparameter Optimization and Meta-Learning

Luca Franceschi, Paolo Frasconi, Saverio Salzo, Riccardo Grazzi, Massimilano Pontil

Intenrnational Conference of Machine Learning, 2018.

PDF Code

AutoML

Far-HO: A Bilevel Programming Package for Hyperparameter Optimization and Meta-Learning

Luca Franceschi, Riccardo Grazzi, Massimiliano Pontil, Saverio Salzo, Paolo Frasconi

AutoML Workshop, ICML 2018

PDF