The Estimation of Time-Varying Graph Signals from Incomplete Observations

Project Coordinator: Assist. Prof. Elif Vural
Project Type: TUBITAK 1001 Scientific and Technological Research Projects Funding Program
Project Budget: 466 000 TL
Project Duration: 30 months
Project Start Date: January 15, 2021
Funded Personnel: 2 PhD Students (Full-Time), 2 MSc Students (Full-Time)

Brief Summary: Many applications nowadays involve the acquisition of data over certain time intervals on platforms such as social networks, communication networks and irregular sensor arrays. Such data types can be modeled as time-varying graph signals. In many practical applications, it is not possible to observe graph signals as a complete data set with no missing samples; hence the need for estimating the unavailable observations of the graph signal using the available ones arises as an important requirement. Among many applications involving time-varying graph signals, some examples are the prediction of how an epidemic will evolve in a society, the estimation of lost data in a sensor network due to sensor failure or communication problems, the prediction of the tendencies of individuals in a social network such as their consumption habits. The purpose of this project is to develop novel methods for the modeling and estimation of time-varying graph signals.