My current research is mainly focused on the development of nonlinear optimization methods with applications to the solution of problems arising from Machine Learning, Deep Learning and Image Processing. I am particularly interested in large- and huge-scale problems involving nonsmooth and/or nonconvex regularization terms such as those enforcing some structured sparsity in the solution.
Evan Murphy (2024-Present), PhD program in Mathematical Sciences at DCU (Ireland)
Seonjae Kang (2024), Master program in Algorithms and Learning at Sorbonne University (France)
Gabriele Sanguin (2024), Master program in Data Science at University of Padua (Italy)
Sep 2024 - Present: “Probabilistic Approaches for Object-Level Analysis of Natural Images”, in collaboration with Vladimir Krylov (DCU) and Evan Murphy (DCU)
Sep 2023 - Present: “Efficient training algorithms for deep learning on traditional and neuromorphic architectures”, UCD STEM Challenge Fund 2023. The project is a collaboration with researchers at UCD School of Computer Science, UCD School of Mechanical and Materials Engineering, and University of Padua.
We are currently focused on the analysis of self-calibration strategies in the training of deep neural networks through the use of bilevel programming.
2023: “Investigating advanced optimization methods for deep neural networks training”, School of Mathematics and Statistics: Collaborative Research Projects Grant Scheme 2023. Collaboration with researchers at University of Turin (Italy).
2019-2021: “Second order methods for optimization methods in machine learning”, Executive programme of cooperation in the field of science and technology between the Italian Republic and the Republic of Serbia for the years 2019-2021, Coordinators: Prof. Stefania Bellavia (University of Florence, Italy), Dr. Natasa Krklec Jerinkic (University of Novi Sad, Serbia)
2020: “Optimization for Machine Learning and Machine Learning for Optimization", INdAM-GNCS Projects, funded by Istituto Nazionale di Alta Matematica, Coordinator: Dr. Federica Porta (University of Modena and Reggio Emilia, Italy)
2019-2021: "VArIational methods and Numerical techniques: sHape Optimization and nonlinear Partial differential EquationS - VAIN-HOPES", 2019 V:ALERE (VAnviteLli pEr la RicErca) Program of the University of Campania "L. Vanvitelli". Coordinator: Prof. Benedetta Pellacci