#7 MLMilan: Fight neural network memorization and survival analysis

Questo è un evento passato

Hanno partecipato 111 persone

Immagine del luogo dell'evento


Dear all,

Milan is getting hotter but a new event from ML Milan is going to come to refresh your minds before the vacation break. It will take place in the beautiful location of Google Milan office and this time we will have two talks from Riccardo Lorenzon and Andrea Marisio.

“Use survival analysis to estimate spare parts market potential”
During the speech Riccardo will go through challenges and opportunities of using customer service data to run survival analysis on product components and subsequently estimate the market potential for spare parts.

Seasonality, censored data, partial observation windows and mixture of distributions are only a few of the technical issues that have to be tackled. The purpose of the talk is to propose a possible approach and brainstorm over potential other solutions.

Riccardo Lorenzon
Riccardo is Lead Data Scientist in Whirlpool EMEA, previously Senior Data Scientist in Data Reply. He holds a Master of Science at Politecnico of Milano in Mathematical Engineering with a focus on Mathematical Models for Decision Making. He specialized in Artificial Intelligence at ETH Zurich.
He has hands-on experience in end-to-end data-driven projects development in various industries, with a focus on manufacturing and retail. He supports the design of Data Strategies, ranging from business cases identification to algorithms and data products development.

“Remember to forget: fight neural network memorization”
In this talk, Andrea will show you a New Theory that cracks open the Black Box of Deep Learning with Information theory: A new idea called the “information bottleneck” is helping to explain the puzzling success of today’s artificial-intelligence algorithms. He
will put down his own research to help you deal with deep learning memorization. Recent works changes the way we must think about generalization. With the help of Yoshua Bengio's ideas and Information theory we will try to open the Black Box of Deep Learning and how to deal with it.

Andrea Marisio
Andrea is a Deep Learning Engineer at Boom Image Studio. In his previous experience he did a master at Università degli Studi dell’Insubria and worked in NTT DATA. Andrea is passionate about learning and teaching... to machines. Guided by his passion he is helping businesses to grow in the new AI era.

Where? Join us on Thursday 11th of July at Google office, in Via Federico Confalonieri, 4. The event will start at 19:00 and will finish around 20:30, with an appetizer offered from Google at the end of the event.

Looking forward to see you all,