Panel: PAN-1: Data Science Education: The Signal Processing Perspective

In the last decade the signal processing community has witnessed a paradigm shift from model-based to data-driven methods. Machine-learning, and specifically deep learning, methodologies are nowadays widely used in all signal processing fields, e.g., audio, speech, image, video, multimedia, multi-modal and multi-sensor processing, to name a few. Many data-driven methods are also incorporating domain-knowledge to improve the problem modeling, especially when computational burden, training data scarceness, and the size of memory are important constraints. Data science, as a research field, emerged from several scientific disciplines, namely mathematics (mainly statistics and optimization), computer science, electrical engineering (mainly signal processing), industrial engineering and information systems. Each of these disciplines suggests an independent teaching program in its core domain with a segment in data science. In recent years, several institutes world-wide are starting to offer dedicated data science teaching programs that can be used for different application areas. We believe that there is a unique signal processing perspective of data science that should be reflected in the education we would like to give to our students. Moreover, we think that we are in the correct time to start defining our needs and inspirations. In this panel, we shall focus on these education aspects and, hopefully, draft a manifesto for a SP-oriented data science curriculum. Discussion Questions How are �??signals�?? defined? Specifically, does a signal always represent an underlying physical phenomenon? Or can it represent a cognitive space in our brains, e.g., semantics? The key question: Is there a unique perspective of signal processing in data science that is different from the viewpoints of other disciplines? If this is indeed the case, can we come up with a clear definition of this perspective? Should data science programs be offered already at undergraduate level studies, or should it be postponed to graduate level studies? We should keep in mind the different education systems in different countries with either a 4-year BSc program or a 3+2 BSc+MSc program. What should be the different roles of the BSc and MSc programs in educating the future DS scientists/engineers?What should be regarded as core undergraduate education in the field? Which parts are mandatory and which can be elective? We can propose the following topics, and open the floor to more ideas: Mandatory: 1) Mathematics; 2) Statistics; 3) Computer skills and algorithms; 4) Signal processing and Machine Learning; 5) Ethics. Elective: Students should specialize in a specific topic(s) from the proposed list and must also select 3-4 application courses. Proposed list of specialization track (that can be changed/extended): 1) Advanced algorithms and optimization; 2) Security and privacy preservation; 3) Data sharing and communication over networks; 4) Applications in diverse fieldsTeaching methodologies: Does DS education require different teaching methodologies than in regular engineering/CS/math education? Do we need to extend the availability of online courses, flipped classes, hands-on, group work, projects? (If time permits) Topics for graduate (MSc and PhD level) studies in the field? (If time permits) Advanced studies are research-oriented, and the list of topics should reflect the research activities of the department. This may include: 1) Audio processing; 2) Image Processing and computer graphics; 3) Machine learning theory; 4) Deep learning; 5) Natural language processing; 6) Multi-modal and multi-sensor processing; 7) Bio-medical processing; 8) Networks and communications