About Me

I recently finished my PhD in Machine Learning at UCL , London. I worked under the supervision of Massimiliano Pontil and Carlo Ciliberto. In November 2020, I started a PostDoc at the Department of Computing at Imperial College .

Research Interests

My main research interest focuses on the interplay of Machine Learning and Optimal Transport, in particular on using optimal transport distances for learning probability measures in supervised or unsupervised settings. I also work a bit on structured prediction and surrogate frameworks. Recently, I started working on more applied topics -targeting Reinforcement Learning for healthcare.

Education

• Research Associate at Imperial College, Department of Computing
• PhD student at UCL, Computer Science department
  • Apr. 2019-July 2019 I worked at CREST , ENSAE ParisTech , under the supervision of Marco Cuturi and Gabriel Peyre’
• MSc. Mathematics, University of Pavia, 2014-2016
  • MSc dissertation: Contraction and regularizing effects for Markov semigroups in Hellinger-Kantorovich spaces
  • Advisor: Giuseppe Savare’
• Exchange Program (Mar. 2015- Aug. 2015) at Humboldt University, Berlin
• BSc. Mathematics, University of Pavia , 2011-2014

You can find my full CV here .

Publications

• GL, M. Pontil, C. Ciliberto, Generalization Properties of Optimal Transport GANs with Latent Distribution Learning , submitted

• S. Cohen, GL , A. Terenin, A. Brandon, M. Deisenroth, Aligning Time Series on Incomparable Spaces , AISTATS 2021

• A. Salim, A. Korba, GL, Wasserstein Proximal Gradient , NeurIPS 2020

• L. Oneto, M. Donini, GL, C. Ciliberto, A. Maurer and M. Pontil, Exploiting MMD and Sinkhorn Divergences for Learning Fair and Transferable Representations , NeurIPS 2020

• GL, S. Salzo, M. Pontil, C.Ciliberto, Sinkhorn Barycenter with Free Support via Frank-Wolfe Algorithm , NeurIPS 2019 (spotlight)

• GL, G. Savare’, Contraction and regularizing properties of heat flows in metric measure spaces , DCDS-Series S, doi: 10.3934/dcdss.2020327

• GL, D. Stamos, M. Pontil, C. Ciliberto, Leveraging Low-Rank Relations Between Surrogate Tasks in Structured Prediction, ICML 2019.

• GL, A. Rudi, M. Pontil, C. Ciliberto, Differential Properties of Sinkhorn Approximations for Learning with Wasserstein Loss , NeurIPS 2018.

News (conferences/workshops I have attended)

• MalGA Seminars , 16 March 2021
• DeepMind CSML Seminar Series , 28 Nov 2019, London, (slides)
• MAGA days 20-21 Nov 2019, Paris (slides)
• POTA workshop Jun. 2019, Cortona (Italy)
• ICML Jun. 2019, Long Beach (poster)
• NeurIPS Dec. 2018, Montreal (poster)
• Optimal Transportation and Applications 2018, Pisa, Italy
• MLSS 2018 , Buenos Aires