The Faculty Exchange Program “Was a Very Fruitful and Successful Experience”
| ||Petia Georgieva, a researcher at the Instituto de Engenharia Eletrotécnica e Telemática de Aveiro (IEETA) and assistant professor at the Department of Electronics Telecommunications and Informatics of the Universidade de Aveiro, spent three and a half months at Carnegie Mellon University working as a visiting professor and researcher, as part of the Faculty Exchange Program of the Carnegie Mellon Portugal Program. |
“I was involved in the course on Machine Learning (10-601) given by Tom Mitchell and was responsible for giving recitations, designing homework and exams. Furthermore, I attended lectures and teaching assistant meetings,” she described. This teaching activity also allowed her to update her lectures on the Machine Learning module given to graduate students of the Ph.D. program on Electrical Engineering of the Universidade de Aveiro.
The research was conducted in collaboration with Fernando De La Torre, a researcher at the Computer Vision Center in the Robotics Institute. “We have worked on a new approach to differentiate cognitive brain states directly from functional Magnetic Resonance Images (fMRI),” she stated. “We applied Robust Principal Component Analysis (RPCA), a theoretical framework for dimensionality reduction developed a few years ago by Fernando, to construct low dimensional linear-subspace representations from the noisy fMRI images for each subject, and afterwards we performed a Gaussian Naive Bayes (GNB) classification,” she added.
Their work represented a significant progress to the area comparatively to previous studies. In the past, cognitive brain states from fMRI were differentiated by transforming the fMRI into a time sequence of voxels from which the brain states are inferred. The work by De La Torre and Georgieva improved the classification rate of a real benchmark fMRI data. Petia Georgieva gave a talk in the Robotics Institute (Joint VASC-CBI Seminar) to present her work on brain imaging.
During her stay at CMU, she worked on and wrote three papers: “Robust Principal Component Analysis for improving cognitive brain states discrimination from fMRI”, accepted for presentation at the IbPRIA 2013, 6th Iberian Conference on Pattern Recognition and Image Analysis; “Bayesian approach for reconstruction of moving brain dipoles,” accepted for presentation at the ICIAR 2013, Int. Conf. on Image Analysis and Recognition; and “A particle filter framework for localization of dynamic EEG sources,” submitted to the journal PlosOne.
Petia Georgieva also attended weekly meetings of the research group on intelligent robots, promoted by Manuela Veloso (CORAL) from the Computer Science Department. She “learned about CoBots - Collaborative Mobile Robots.” Furthermore, the researcher participated in meetings of the research group of Jelena Kovačević of the Center for Bioimage Informatics at CMU, as well as in Ph.D. thesis proposal presentations and Ph.D. defenses, and in seminars and talks given by technological and scientific leaders from all over the world.
According to Petia Georgieva, her stay at CMU was “very fruitful and successful” and she hopes to remain involved in the Carnegie Mellon Portugal Program, for example through the co-supervision of doctoral students with professors from Carnegie Mellon.