Simone Marullo (University of Siena)

When: May 19, 2021 – 11:00 – 11:45 AM

Where: Google meet link

Description

Two papers by Simone Marullo, Matteo Tiezzi, Marco Gori and Stefano Melacci will be presented:
– Friendly Training: Neural Networks Can Adapt Data To Make Learning Easier
– Being Friends Instead of Adversaries: Deep Networks Learn from Data Simplified by Other Networks

A novel training procedure named Friendly Training will be presented, based on the idea of altering the training examples to implement a more gradual learning process. Two possible instances of this training method for neural networks will be exposed, the former is based on direct gradient-based optimization while the latter is based on an auxiliary network. The relationship between our proposal and existing work will be discussed along with some experimental results.