Resource Efficient Machine Learning
This project aims to develop efficient machine learning solutions by blending statistical and algorithmic principles to design scalable learning machines.
LCSL
The Laboratory for Computational and Statistical Learning (LCSL) is one of the research units within MaLGa, the machine learning center of the University of Genova. LCSL focuses on the development of efficient and reliable machine learning algorithms blending tools from statistics, optimization, and regularization theory.
Our goal is to develop theoretically grounded and practical machine learning algorithms. In collaboration with other MaLGa units and external collaborators we work on foundational aspects of Machine Learning (CHARML) and also a number of applications including vision (MLV), robotics, and biological behavior (PiMLB).
Research topics
Optimization for Machine Learning
We aim to develop flexible and scalable optimization algorithms for machine learning, emphasizing robustness and adaptivity to tackle diverse and complex data-driven problems effectively.
Structured Machine Learning
We tackle the challenge of handling structured data in machine learning, addressing various domains and tasks such as text structure inference, subgraph identification, and time series forecasting.
Explainable and Reproducible Machine Learning
Machine Learning for Health
People
Annalisa
BarlaSimone
Di MarinoCesare
MolinariLorenzo
RosascoSilvia
VillaFrancesco
MontagnaEdoardo
CaldarelliEmilia
MagnaniArnaud
WatusadisiIlaria
StanzaniMarco
LetiziaAndrea
Della Vec…Hippolyte
LabarrièreShuo
HuangEmanuele
NaldiCheik
TraoréPietro
ZerbettoStefano
RaveraRayan
AutonesJoachim
Bona-Pellis…
Past People
People
Elena Milano | 2024 | Research Scholar | Content strategy for ethical-AI communication
Ettore Fincato | 2023 → 2024 | PhD student
Cheik Traoré | 2020 → 2024 | PhD student | Optimization
Marco Rando | 2020 → 2024 | PhD student
Cristian Jesus Vega Cereno | 2021 → 2024 | PhD student | Optimization
Jonathan Chirinos Rodriguez | 2024 | PhD student
Antoine Chatalic | 2021 → 2024 | Post-doctoral fellow
Rosanna Turrisi | 2023 → 2024 | Post-doctoral fellow | Machine Learning
Elisa Maiettini | 2021 → 2023 | Post-doctoral fellow
Nicolas Schreuder | 2020 → 2023 | Post-doctoral fellow
Vassilis Apidopoulos | 2019 → 2023 | Post-doctoral fellow
Gabriele Bortolai | 2023 | Student
Ilaria Stanzani | 2023 | Research Scholar | Machine Learning
Federico Ceola | 2020 → 2023 | PhD student
Andrea Maracani | 2020 → 2023 | PhD student
Paolo Didier Alfano | 2022 → 2023 | Research Scholar | Machine Learning and vision
Andrea Della Vecchia | 2022 → 2023 | Research Scholar | Machine Learning
Silvia Vertemati | 2023 | Student
Daniele Giampaoli | 2022 | Student | Structured Learning
Sophie Langer | 2022 | Post-doctoral fellow
Research projects
Most recent LCSL research projects
Integrating network science and explainable AI to forecast the impact of climate change on infectious disease risk
Theme Machine Learning for HealthUnit LCSLLeading scientist Annalisa BarlaRole within the project Head of unitDuration 2023 - 2025Sponsors MIUR ¦ PRINTotal funding 250 k€- ML4IP
Reliable and efficient machine learning for inverse problems
Unit LCSLLeading scientist Lorenzo RosascoRole within the project Principal InvestigatorDuration 2022 - 2025Sponsors FARE 2020Total funding 300 k€ - SLING
Efficient algorithms for sustainalble machine learning
Unit LCSLLeading scientist Lorenzo RosascoRole within the project Principal InvestigatorDuration 2019 - 2024Sponsors EU | ERC CoGTotal funding 1.98 M€ Traning data driven experts in optimization
Unit LCSLLeading scientist Silvia VillaRole within the project Principal InvestigatorDuration 2020 - 2024Sponsors EU | MSCA ITN ETNTotal funding 3.77 M€
Publications
Most recent LCSL publications
| Title | Year | Author | Venue |
|---|---|---|---|
| NLP-based tools for localization of the epileptogenic zone in patients with drug-resistant focal epilepsy | 2024 | Mora Sara; Turrisi Rosanna; Chiarella Lorenzo; Consales Alessandro; Tassi Laura; Mai Roberto; Nobili Lino; Barla Annalisa; Arnulfo Gabriele | SCIENTIFIC REPORTS |
| RESPRECT: Speeding-up Multi-Fingered Grasping With Residual Reinforcement Learning | 2024 | Ceola Federico; Rosasco Lorenzo; Natale Lorenzo | IEEE ROBOTICS AND AUTOMATION LETTERS |
| Iterative regularization for low complexity regularizers | 2024 | Molinari C.; Massias M.; Rosasco L.; Villa S. | NUMERISCHE MATHEMATIK |
| Assumption violations in causal discovery and the robustness of score matching | 2024 | Montagna Francesco; Mastakouri Atalanti A.; Eulig Elias; Noceti Nicoletta; Rosasco Lorenzo; Janzing Dominik; Aragam Bryon; Locatello Francesco | Advances in Neural Information Processing Systems |
| Top-tuning: A study on transfer learning for an efficient alternative to fine tuning for image classification with fast kernel methods | 2024 | Alfano P. D.; Pastore V. P.; Rosasco L.; Odone F. | IMAGE AND VISION COMPUTING |
Open data & software
- LCSL
CRiSP for Misspecified Robot Model
Gian Maria Marconi, Raffaello Camoriano, Lorenzo Rosasco, and Carlo CilibertoCode accompanying the paper “Structured Prediction for CRiSP Inverse Kinematics Learning with Misspecified Robot Models”
- LCSL
Iterreg: Iterative Regularization For Linear Models
C. Molinari, M. Massias, L. Rosasco and S. VillaIterreg is a scikit-learn compatible python package to perform iterative regularization of linear models. It implements the algorithm of “Iterative regularization for convex regularizers” by C. Molinari, M. Massias, L. Rosasco and S. Villa, AISTATS 2021.
- LCSL
Falkon 2.0: Scaling Nyström KRR to Billions of Points
Giacomo Meanti, Luigi Carratino, Lorenzo Rosasco, Alessandro RudiPython package implementing the Falkon algorithm for large-scale, approximate kernel ridge regression. The implementation is based on PyTorch and runs on CPU and GPU.
- LCSL
Batch-BKB: Near-linear time, no regret Bayesian optimization
Calandriello Daniele, Luigi Carratino, Alessandro Lazaric, Michal Valko and Lorenzo RosascoPython code implementing Batch-BKB: the first Bayesian optimization (a.k.a. Gaussian process or bandit optimization) algorithm that is both provably no-regret and guaranteed to run in near-linear time time.