Introduction to Professor Santiago Segarra and His Lab at Rice University
1. Could you briefly introduce yourself (and your University/Lab)?
Originally from Argentina, I came to the US for my graduate studies in 2011. I did my PhD at the University of Pennsylvania and a postdoc at MIT. Since 2018 I am an Assistant Professor in Electrical and Computer Engineering (and Computer Science, by courtesy) at Rice University. I work at the interface of signal processing, graph theory, statistics, and machine learning. In particular, I have been working on graph signal processing since 2014 and, more recently, on machine learning for graphs and graph data.
2. What have been your most significant research contributions up to now?
I’d say that these are within the area of graph signal processing. In particular, I contributed on some of the first papers in the topics of sampling and reconstruction of graph signals, graph filter design, blind deconvolution of graph filters, graph stationary processes, and network topology inference from the observation of graph signals.
3. What problems in your research field deserve more attention (or what problems will you like to solve) in the next few years, and why?
Over the next few years I will focus mainly on three broad directions. The first one is about signal processing and machine learning on higher-order networks, i.e., relational structures that go beyond pairwise interactions and include group-wide interactions. This would drastically expand the applicability of graph tools into more general structures, posing interesting challenges both from theoretical and applied perspectives. The second direction is the development and theory of graph (and higher-order) neural networks. Undoubtedly, these graph-based trainable structures have shown lots of promise on several problems of interest during the last few years, but they remain quite elusive from a theoretical perspective. By leveraging insights and concepts from graph signal processing, we would be able to provide a unique view into the challenging problem of truly understanding graph neural networks. My last direction is concerned with real-world applications of these tools. In particular, I am (and will be) working on applications in the fields of wireless communications, metagenomics, seismology, and social networks, among others.
4. What advice would you like to give to the young generation of researchers/engineers?
Well, I am probably not old enough to give advice to “younger” generations but maybe two suggestions that I often repeat to my students could resonate with others out there. First, I encourage researchers to have a portfolio of projects on which you can work simultaneously. This not only provides technical breadth but also reduces the risk associated with the uncertainty of what will be adopted by the community and mitigates the frustration if a specific project is not going as expected. Second, even when working on theoretical projects, think about the real-world motivation of the model under study. This will certainly help during the review process and, more importantly, will facilitate true adoption down the road.