The Computational Mathematics and Big Data Sets Analytics team's seminar
On 29.10.2024 (Tuesday) at 11:30 am
in room 304 (A3-A4 building)
Professor Hernando Ombao
(King Abdullah University of Science and Technology, Saudi Arabia)
will present a talk:
„Robust Canonical Coherence”
You are cordially invited!
Abstract: In this talk, we consider the problem of characterizing a robust ``global” dependence between two brain regions. This work is driven by experiments to investigate dependence between two cortical regions and identify difference in two brain states. A common approach to explore dependence between two multivariate signals is via canonical correlation analysis (CCA) which is limited to capturing linear associations and is sensitive to outliers. These limitations are crucial because brain network connectivity is likely to be more complex than linear and that brain signals may exhibit outlier observations. To overcome these limitations, we develop a robust method, Kendall canonical coherence (KenCoh), for learning frequency-specific monotonic connectivity structure among brain oscillations extracted from the observed signals. Furthermore, we propose using a permutation test based on KenCoh for testing the difference in brain network connectivity across two different states. Our simulations demonstrate that KenCoh is competitive to the traditional variance-covariance estimator, especially for heavy-tailed distribution. The KenCoh method yielded interesting results when applied to brain signals.
This is joint work with Mara Talento (Ph.D. student) and Dr. Sarbojit Roy (Post-Doctoral Scholar) in the Biostatistics Group at KAUST.