- - machine learning methods for signal processing
- - statistical signal processing
- - graphical networks
- - nonlinear sequential estimation
- - sensor networks
During my postgraduate studies in Graz University of Technology, my research focused on modeling and prediction the temporal variations of wireless MIMO channels. It is known that mobile wireless channels are subject to often extreme temporal variations, also known as fading. Due to multipath and Doppler effects these variations are in general nonlinear. In my thesis I aimed to study how these dynamical changes can be modeled, which led me to the definition of the channel ``hypermodel'' that deterministically represents channel variations. By combining high resolution model-based estimation algorithms, tracking, and modern machine learning tools, I proposed a framework to learn channel hypermodels sequentially from the measured channel impulse responses and use them for channel forecasting.
Currently I am involved in research of machine learning algorithms within the context of wireless systems, in particular within the context of intelligent (cognitive) ad hoc wireless sensor networks. I am interested in performance analysis and advancement of state-of-the-art distributed learning algorithms under communication constraints, e.g,. bandwidth, or power constraints, and dynamical network structure perturbations. I also investigate potential gains of cross-layer designs of distributed learning algorithms.
A list of my publications is available here.