Lapo Faggi (University of Siena)

When: Jun 1st, 2022 – 11:00 – 11:45 AM

Where: Google meet link

Description

Neural dynamic in temporal environments
Learning in a continual manner is one of the main challenges that the machine learning community is currently facing. The importance of the problem can be readily understood as soon as we consider settings where an agent is supposed to learn through an online interaction with a data stream, rather than operating offline on previously prepared data collections. In the last few years many efforts have been spent in proposing both models and algorithms to let machines learn in this setting, usually re-adapting classic statistical approaches, but the problem still remains extremely challenging. In this talk, I will present some alternative ideas, developed by researchers of our lab, to define an appropriate weights temporal dynamic in neural nets. In particular, we will analyze the dynamic resulting from classical variational approaches and the one originating from an optimal control perspective. We will compare those methods, underlining pros and cons of each alternative.