
Dr. Leonardo Galli
I am a postdoctoral researcher at the Chair working under the guidance of
Prof. Holger Rauhut. I am now working on stochastic line search methods for training deep neural networks (see ROGER) together with him and Prof. Mark Schmidt.
I obtained my B.S., M.S. and Ph.D. degrees from
University of Florence respectively in 2013, 2016 and 2020. During my studies
I had a few international collaborations. In 2015, I spent a research period at
University of Würzburg where I worked on generalized Nash equilibrium
problems together with professor Prof. Christian Kanzow. Starting from 2018,
I collaborated with professor Prof. Chih-Jen Lin on truncated Newton methods
for linear classification. In particular, we aimed at updating LIBLINEAR, one of
the most widely used and cited linear classification packages. Thanks to this
project, I could visit professor Lin for a research period at UCLA in 2019 and
at National Taiwan University in 2020. The project has ended in 2020, when
a faster version of the LIBLINEAR software was released.
Research interests:
My major research areas include optimization methods, machine learningand operational research. In particular, together with my former tutors Prof. Marco Sciandrone and Prof. Fabio Schoen, I deeply explored the
fields of nonmonotone techniques and prescriptive analytics.
I am now working on line search methods for deep learning.