Efstratios Tsoukanis joined the CGU Institute of Mathematical Sciences as a Research Assistant Professor in 2024. His research focuses on applied harmonic analysis and group representation theory, with specific interests in invariant machine learning and the phase retrieval problem. He has conducted extensive explorations in constructing group invariant embeddings into low-dimensional Euclidean spaces, delving into their injectivity and stability.
Tsoukanis earned his BS and MS degrees from the University of Crete and his PhD from the University of Maryland under Dr. Radu Balan. His dissertation was on G-invariant representations using co-orbits, an embedding scheme allowing for efficient training of machine learning models on graphs that are invariant to node labeling. He is currently working with IMS Distinguished Research Professor Hrushikesh Mhaskar on research in computational harmonic analysis and machine learning, funded by the US Office of Naval Research.
R. Balan, E. Tsoukanis, “Relationships between the Phase Retrieval Problem and Permutation Invariant Embeddings,” Sampling Theory and Applications Conference (SAMPTA 2023), Yale University, arXiv:2306.13111.
R. Balan, E. Tsoukanis, “G-Invariant Representations using Coorbits: Bi-Lipschitz Properties,” arXiv:2308.11784 (2023).
R. Balan, E. Tsoukanis, “G-Invariant Representations using Coorbits: Injectivity properties,” arXiv:2310.16365 (2023).