Today it is my turn to tell something concerning my mission out from Florence. I have been for the whole week to the very nice ancient town of Veroli, 100km from Rome. Veroli is really an interesting place, an ancient town, with many interesting things to see, very good food, really nice people, ….
But we had been there not just for those very good reasons… From June 30th to July 6th the Second Edition of the OBA summer school took place there.
This is an international event offering top level courses to PhD’s and postdoc’s in the hot field of machine learning and optimization.
All of our lab took part to this event: Marco Sciandrone, who is the main organizer, the scientific director of the School (and the only one of us who was born in Veroli); moreover, all in some way involved in the organization, we had Tommaso Levato, Leonardo Galli, Gulio Galvan, Luca Tigli, Matteo Lapucci, Alessio Sortino, Leonardo Di Gangi, Guido Cocchi, Roberta Rossi, Enrico Civitelli.
The event was (as in the first edition) a great success; more than 70 participants from all over the world followed 4 series of top-level lectures by world reknown experts in the field:
- Nicolò Cesa-Bianchi, University of Milan (Italy), author of many fundamental papers with thousands of citation (see the scholar page), who gave a series of lectures on on line learning and algorithms, like On line Stochastic Gradient, Mirror Descent, to name afew
- Julien Mairal, Inria Grenoble, head of the Thoth team. Julien is a top scientist who, among many other achievements, received the “test of time award at ICML 2019”. Here are the video, and the slides. He gave really stimulating lectures on Stochastic Optimization for large scale machine learning
- Miguel Á. Carreira-Perpiñán, Professor of Computer Science at University of California, Merced. Miguel gave us a series of lectures on how to train in an “efficient” way deep networks and how to train a decision tree in an optimal way; many of his research achievements are connected with the clever use of penalty methods.
- Andrew D. Bagdanov, Professor of Computer Science at DINFO, University of Florence. Andy, a colleague from my department, is a world known in Computer Vision. He gave a series of very wide and illuminating lectures on deep learning, convolutional neural networks, GAN’s and beyond.
As it became a tradition, Friday afternoon we had a session of presentations from a few selected students. Nine students gave a 15-minutes presentation of their current research on the topic; many of them were indeed very interesting and stimulating. The award Committee had quite a difficult time in deciding who deserved the final prize.
And, in fact, eventually the OBA 2019 Award went, ex aequo, to 2 presentations, sharing the prize:
- Nikita Doikov “Randomized Block Cubic Newton Method”
- Elisa Riccietti “Solution of PDEs by multilevel artificial neural networks”