Last few decades, research on convex optimization has led to theoretically sound as well as practically efficient tools for solving a wide spectrum of optimization, estimation and design problems. A main outcome is that this toolbox has helped to separate algorithmic questions from conceptual matters of the optimality task at hand: availability of reliable software implementations allows the researcher to focus on conceptual issues, while at the same time enabling to verify ideas in practical case studies. This approach came first up in research on identification in nonlinear systems. More recently, it is also found to push the limits of identification of linear systems.

Participants of this research project:

• Senior: Kristiaan Pelckmans
• Ph.D. student: Liang Dai

Published results include:

• Pelckmans K., Goethals I, J.A.K. Suykens, B. De Moor (2005). On model complexity control in identification of Hammerstein Systems, Conference on Decisions and Control (CDC-ECC), Sevilla, Spain, pp.1-8.
• Pelckmans K., J.A.K. Suykens, B. De Moor (2007). Convex Optimization for the Design of Learning Machines, 15th European Symposium on Artificial Neural Networks (ESANN2007), Bruges, Belgium, pp.1-8.
• Pelckmans K. (2010) On the Identification of Monotone Wiener Systems, In the 49th IEEE Conference on Decision and Control (CDC2010), Atlanta, US, Dec. 2010, pp. 7208-7213.
• Babu P., Pelckmans K., Stoica P., Li J. (2010). Linear Systems, Sparse Solutions, and Sudoku, IEEE Signal Processing Letters, vol. 17, no 1