IIT Home Page CNR Home Page

On Network Connectivity for Distributed Machine Learning

Many learning problems are formulated as minimization of some loss function on a training set of examples. Distributed gradient methods on a cluster are often used to this purpose. In this talk we discuss how the variability of task execution times at cluster nodes affects the system throughput. In particular, a simple but accurate model allows us to quantify how the time to solve the minimization problem depends on the network of information exchanges among the nodes. Interestingly, we show that, even when communication overhead may be neglected, the clique is not necessarily the most effective topology, as commonly assumed in previous works.

Dal 22/02/2019-10.00 al 22/02/2019-11.00 , Area della Ricerca di Pisa (CNR), Aula A32

Speaker: Giovanni Neglia

Responsabile: Andrea Passarella


Note: Biografia: Giovanni Neglia received the master’s degree in electronic engineering and the PhD degree in telecommunications from the University of Palermo, Italy, in 2001 and 2005, respectively. He has been a researcher at Inria, Sophia Antipolis, France, since September 2008. In 2005, he was a research scholar with the University of Massachusetts, Amherst, visiting the Computer Networks Research Group. Before joining Inria, he was a post-doctorate with the University of Palermo and an external scientific advisor with the Maestro Team at Inria. His research is focused on modeling and performance evaluation of networks.