Discovering the number of latent components in tensor decomposition
Tensor decomposition has been shown, time and time again, to be an effective tool in multi-aspect data mining, especially in exploratory applications where the interest is in discovering hidden interpretable structure from the data.
In such exploratory applications, the number of such hidden structures is of utmost importance, since incorrect selection may imply the discovery of noisy artifacts that do not really represent a meaningful pattern. Albeit extremely important, selection of this number of latent factors, also known as low-rank, is very hard, and in most cases, practitioners and researchers resort to ad-hoc trial-and-error, or assume that somehow this number is known or is given via domain expertise.
For this reason, in this talk we will briefly go over a few heuristics that have been used to achieve the desired low-rank, and then we will talk about a novel method of ours that has shown promising results in multiple real-world datasets.
About the Speaker
Yorgos Tsitsikas earned his Diploma in Electrical and Computer Engineering from Technical University of Crete in 2017, where he received the Limmat Foundation award for being among the students with the top 3 graduation grades.
He immediately started his Ph.D. program at University of California, Riverside, where he focuses on tensor rank estimation and tensor decompositions, and their relation to real world applications. His research interests also span broader topics such as machine learning, optimization, information theory, the human brain, and any other topic that could prove useful in advancing artificial intelligence.
His most recent work has been published in International Conference on Data Mining (SDM20), which was next nominated and accepted for publication in the special issue on "Best of SDM 2020" of Big Data Journal.
Currently, he is doing his summer internship at Los Alamos National Laboratory where he joined the Computational Earth Science (EES-16) Group in the Earth and Environment Science Division. His work there is focusing on the theoretical development of tensor decompositions with physics constraints.