Efficient Machine Learning for Dependable Networks

This project addresses challenges in applying machine learning to metrics from video delivery networks, particularly within Software Defined Networks (SDN).

To engineer future network systems with high service dependability, the project focuses on three key objectives:

1. Mapping the performance of underlying network switches to delivered services.

2. Incorporating awareness of factors such as device load, network congestion, and throughput into the learning algorithms.

3. Ensuring fast learning to promptly utilize insights.

Predictive monitoring through optimizing historical data usage and selecting appropriate data subsets is a key aspect. The project emphasizes "Dependable Networks and Customized Networks" and "Data-driven Optimization and Management" of programmable networks.

Building on recent work in Sketching for fast Quality of Delivery prediction, the project investigates learning algorithms for dynamic networks. Participants will gain practical experience by developing research testbeds and implementing fast learning algorithms.