Pooling Information Theory and Signal Processing into Networks

שלחו לחבר
Alejandro Cohen, Massachusetts Institute of Technology (MIT)
BIU Engineering Building 1103, Room 329

In advanced systems and applications, efficient task-oriented representation and acquisition of information are critical to allow new and future complex solutions. In recent years, it has been shown how data-driven methods can allow tremendous applications in many domains of science, industry, engineering, and beyond. These new applications, e.g., streaming, autonomous vehicles, smart cities, IoT, learning applications, etc., already have a significant impact on our society and could potentially change the lives of millions in the future. However, these solutions demand the usage of a massive amount of readily available data that can inform the design and operation of such applications. The data may be captured from several co-located and distributed sources and may be used to perform specific tasks, e.g., learning models, computing a function of the data, or informing a data-aided decision, with security and privacy constraints. This increased demand for network connectivity and high data rates necessitates efficient utilization of all possible resources via heterogeneous networks. In this talk, I will show how we can significantly improve the way we acquire, represent and transmit information by dragging Information Theory and Signal Processing into networks. In particular, I will present the following directions: reliable data acquisition over networks with delay and throughput guarantees, efficient data acquisition methods for tasks, security and privacy in advanced systems and computation in networks.