Postdoc • Penn
ndf96@seas.upenn.edu
I am a postdoctoral researcher at the University of Pennsylvania, where I work under the supervision of Prof. George Pappas.
My research. I work at the interface of communication networks, controls and machine learning in the area of distributed intelligence: where multiple entities or agents (i) take decisions or (ii) learn how to take decisions or (iii) try to gather some understanding of the world - in a decentralized or distributed manner, and/or by communicating and transmitting information. In particular, I study the role of information exchange and transmission for coordination in distributed systems, and design protocols that let entities do more with less information. Recently, I have applied multi-agent intelligence to real-time environmental monitoring with marine robot teams, an exciting and challenging real-world problem. Check out our recent project on mapping the Douro river plume with underwater autonomous vehicles and multi-agent reinforcement learning.
My PhD. I obtained my PhD from the University of Padova, where I was advised by Prof. Luca Schenato and co-advised by Prof. Michele Rossi. My dissertation focused on studying designs and properties of frameworks in which multiple agents cooperate to solve a machine learning (ML) or reinforcement learning (RL) problem by communicating with a central coordinator, in what is called the Federated Learning (FL) paradigm. Some of the research questions I tried to answer have been: can we get superlinear convergence in FL in a communication-efficient way? Is cooperation beneficial in federated RL when communication is subject to constraints? The typical fields of application of these studies are, for example, in large-scale distributed ML, FL over wireless networks, ML and RL algorithms trained from decentralized data at the network edge and 6G systems.
