Member in the Spotlight: Sergios Theodoridis

In this series, we aim to introduce senior society members and other experts of the signal processing field. This month, we are happy to introduce Prof. Sergios Theodoridis, Professor of Signal Processing and Machine Learning in the Department of Informatics and Telecommunications at the University of Athens.

When did you first come into touch with signal processing? What was your motivation of following a career in this domain?

My first degree was in Physics. Although I loved physics, and I still do (reading books for the wide audience about cosmology is a hobby of mine), while I was in the University I wanted to do something more application-oriented and work in Industry. So, I decided to take an MSc course in communications related to microwaves, waveguides and antennas. Such a course was as close to Physics as I could find at that time. Among the electives, it was a course on random signals and communication theory. It was the first time that I heard that such a discipline existed! I said, let me try and see what this is about. As I said to a class-mate of mine at that time: “It sounds very mystical”. The suggested reference was Lathi’ s book. One of the best books I have ever read in the area. I still look at it, from time to time.  After that course, nothing was the same as before. I decided to learn more on signal processing. Moreover, I decided to stay on for a Ph.D and postpone my going to work in Industry for later on. So, basically, my career, as I see it now in retrospect, is due to an instance of curiosity. If I had not selected that course, may be now I would be doing something else, probably not in the University.

What was the first signal processing algorithm you ever implemented? In which context was it used?

My Ph.D thesis was on maximum entropy spectral analysis. It was funded by the UK ministry of defence. This had two sides. The good thing was that I had a good scholarship to live and study, but the bad thing was that the major parts of the thesis were never published, since it was decided to be considered classified. Things were different at that time. In the context of my Ph.D, my first signal processing algorithms, implemented in FORTRAN, were the FFT and the AR spectral analyser. Soon after, both algorithms were implemented in Assembly, as part of a bigger package for radar tracking and detection. The Assembly code was implemented on a PDP-9 computer. I still have the print outs as a souvenir. I have to say that, I enjoyed programming in Assembly.  The other algorithm, which I also implemented at that time in Assembly, was the LMS in order to perform spectral analysis in an adaptive mode of operation.

What are your current research interests in the signal processing field and how these fit in the more general SP research trends?

My current interests revolve around the area where Signal Processing and Machine Learning meet. As a matter of fact, these two disciplines are naturally fused together more and more as time goes by. Machine learning is about “learning from data” and then making predictions. Well, I guess that much of the signal processing research these days is about learning from data. Even classical powerful transforms, such as Fourier, wavelets, etc., which are data-independent, have been moved out of the center of the research happening. The new celebrities are now sparsity, dimensionality reduction, dictionary learning, online learning, massive MIMO, big data, Bayesian learning. All these areas have at their heart a learning mechanism to unveil hidden regularities in the data, which are then modeled and exploited for processing and analysis. Signal processing has now become a core technology that cuts across a number of disciplines. Only a couple of decades ago, strong links of signal processing were identified with a relatively few areas  such as communications, control and seismic sounding. Now, it is hard to think of a scientific field where techniques, which are studied and  developed within our SP community, are not applied and used. Smart grids, social networks, bioinformatics, medical applications, robotics and information retrieval in its modern facet, where audio-music-image-video are involved, are typical examples.

It is true to say that we, in the SP community, are not just doing “processing”. We are also doing “learning” and “analysis”. Moreover, we are not dealing only with what  was traditionally known as “signals”. We are now dealing, more and more, with “information”, in general. I guess, it is difficult to call a DNA sequence a signal (although, strictly speaking, one could do it, yet it would be difficult to be understood within other communities). The same is also true for data related to social networks.

Could you introduce an important state-of-the-art research issue (or technology) in this field?

I am going to say the obvious: Everything that is related to big data, graph-based signal processing, complex systems and networks is part of the state-of-the art palette.

From your experience, is there something the signal processing society can learn from other societies? 

I do not like the word learn. The main issue here is that scientists working in different societies have these days much more in common than they used to have a couple of decades ago. In analogy to what happened with the rest of the world, it has happened to science, too. Nowadays, barriers/borders are a bit difficult to understand and justify. No doubt, every scientific society has a core of problems and a way of approaching these problems that differentiate it from others. However, there are many more problems, which are studied and are common in various scientific communities. Take for example the topic of text-retrieval. Maybe fifteen years ago, it was a core problem in computer science.  Once text was extended to include other modalities, as mentioned before, the corresponding borders between computer science and signal processing are hard to define. The same is true for smart grids. The borders between power engineering and signal processing have also become inconspicuous. Members of our SP society must publish in journals of other societies (this spreads the news concerning our activities around and makes our work known wider). However, at the same time, we must open up to members of other societies and attract them to come to our conferences and publish in our journals. This has happened in the past with the Comms and Control Societies.  It is time to open up to other societies too.

Having said that, an important issue pops in, which we have to “learn”. Not from other societies but from reality. Does the name of our society “Signal Processing” and of our flagship journal “Transactions on Signal Processing”, of which I have the honor to be currently its Editor-in-Chief, portray really what most of our members do? My feeling is that the answer is NO. So, we may have to rethink about redefining our “brand” names. I feel “processing” does not do justice and does not portray what our members do and are active on; and, yes,  we do much more than processing. Some may think of whether the name makes much difference or it is the work that really matters. I am afraid that it is amazing how much, still, a name means to an individual. To open up to other societies, it is important to have a name that really puts a stamp on what we do and who we are. We have to adapt to changes and, after all, adaptivity is a notion at the heart of our research and education. We have to respect the past and our roots. However, we have to look at the future and listen to its messages.

What would be your advice to a new PhD student who wants to start a career in signal processing? 

Listen to her/his own interests and let curiosity be the driving force in her/his research. Make the word “why” the most precious one in the vocabulary.

Brief biography

Sergios Theodoridis is currently Professor of Signal Processing and Communications in the Department of Informatics and Telecommunications at the University of Athens. His research interests lie in the areas of Adaptive Algorithms and Communications, Machine Learning and Pattern Recognition, Signal Processing for Audio Processing and Retrieval. He is the co-editor of the book “Efficient Algorithms for Signal Processing and System Identification”, Prentice Hall 1993, the co-author of the best selling book “Pattern Recognition”, Academic Press, 4th ed. 2008, the co-author of the book “Introduction to Pattern Recognition: A MATLAB Approach”, Academic Press, 2009, and the co-author of three books in Greek, two of them for the Greek Open University. He is the co-author of seven papers that have received Best Paper Awards including the 2014 IEEE Signal Processing Magazine best paper award and the 2009 IEEE Computational Intelligence Society Transactions on Neural Networks Outstanding Paper Award. He has served as an IEEE Signal Processing Society Distinguished Lecturer. He was the general chairman of EUSIPCO-98, the Technical Program co-chair for ISCAS-2006 and co-chairman and co-founder of CIP-2008 and co-chairman of CIP-2010. He has served as President of the European Association for Signal Processing (EURASIP), as member of the Board of Governors for the IEEE CAS Society, and as member of the Board of Governors (Member-at-Large) of the IEEE SP Society. He has served as a member of the Greek National Council for Research and Technology and he was Chairman of the SP advisory committee for the Edinburgh Research Partnership (ERP). He has served as vice chairman of the Greek Pedagogical Institute and he was for four years member of the Board of Directors of COSMOTE (the Greek mobile phone operating company). He is Fellow of IET, a Corresponding Fellow of RSE, a Fellow of EURASIP, and a Fellow of IEEE.

Short Note from eNewsletter's editorial staff

On behalf of the Newsletter's editorial staff, we would like to thank Prof. Theodoridis for the interview and also take the opportunity to congratulate him for two recent awards related to the IEEE Signal Processing Society (both received in 2015): the IEEE Signal Processing Magazine Best Paper Award (together with Konstantinos Slavakis and Isao Yamada) for "Adaptive Learning in a World of Projections: A unifying framework for linear and nonlinear classification and regression tasks", and the Education Award "for sustained contributions to education in the area of machine learning for signal processing."