Public defence in Signal Processing and Data Analytics, M.Sc.(Tech.) Xinjue Wang

The title of the thesis: Advances in Robust Signal Processing and Applications

Thesis defender: Xinjue Wang
Opponents: Prof. Bhavani Shankar Mysore, University of Luxembourg, and Prof. Paolo Di Lorenzo, Sapienza University of Rome, Italy 
Custos: Prof. Esa Ollila, Aalto University School of Electrical Engineering

Modern technologies, such as the Internet of Things (IoT), are becoming increasingly complex and vital to our daily lives. However, these systems often operate in messy, uncertain environments where data can be noisy or incomplete. The doctoral thesis of Xinjue Wang addresses a critical challenge in this field: how to ensure that signal processing and machine learning systems remain reliable and efficient even when facing unexpected disruptions. This research is highly relevant as it provides the mathematical foundation needed to build the next generation of stable and resilient digital infrastructure.

The study tackles three specific problems that often compromise system performance. First, it investigates Graph Convolutional Neural Networks (GCNNs)—a powerful type of AI—and provides new theoretical proofs guaranteeing that they can remain stable even when their data structures change. Second, the research develops new algorithms for detecting device signals in communication systems. These new methods use robust mathematical functions to filter out complex, "heavy-tailed" noise, performing significantly better than traditional approaches. Third, a new framework was introduced for modeling high-dimensional data (tensors), making complex data easier to interpret and analyze.

The results of this thesis provide a suite of novel tools that improve the stability and resilience of signal processing systems. These findings can be applied to design more reliable communication networks, robust sensor systems, and efficient data analysis models. In conclusion, this work advances the theory of robust signal processing, ensuring that our technological systems can function reliably even in the face of real-world uncertainty and interference.

Thesis available for public display 7 days prior to the defence at Aaltodoc.