TGaussian process regression and
recurrent
neural networks for fMRI image classification
Emanuele Olivetti, Sriharsha Veeramachaneni, Diego Sona
Abstract:
We describe our approach for fMRI image classification using Gaussian process regression (GPR) and recurrent neural networks (RNN). The feature attributes were extracted from the 3D volume images by spatially clustering the voxels with high mutual information with respect to the feature ratings. Each cluster mask corresponds to one feature attribute for a volume image and is calculated as the mean of all the voxels in the cluster mask for the particular image. The prediction for movie 3 for a subject is performed using movies 1 and 2 (blanks removed) from the same subject as training using two classification schemes – GPR and RNN.
Emanuele Olivetti received his master's degree in physics from Universita' di Trento, Italy in 2000. He's currently a Ph.D. student in information and communication technologies at the same university and jointly at the Istituto per la Ricerca Scientifica e Tecnologica, Trento, Italy. He is working in the area of machine learning and in particular on active feature sampling techniques. He is also a software developer and free software promoter.web-page: http://sra.itc.it/people/olivetti
Diego Sona received the Laurea degree in 1996 and the Ph.D. degree in 2002 in computer science from the University of Pisa, Italy. He is currently a research scientist at the Automated Reasoning Systems division of the Institute for Scientific and Technological Research (ITC-irst), Trento, Italy. His research interests include neural networks, pattern recognition, information retrieval, relational machine learning and web mining.web-page: http://sra.itc.it/people/sona
Sriharsha Veeramachaneni received his Ph.D. in computer engineering from Rensselaer Polytechnic Institute, Troy, New York in 2002. He is currently a research scientist at the Istituto per la Ricerca Scientifica e Tecnologica, Trento, Italy. His research interests include statistical pattern recognition, machine learning, and information theory.web-page: http://sra.itc.it/people/sriharsha
ITC/IRST (Center for Scientific and Technological Research) is a public research Institute located in Trento in northern Italy. More than 300 researchers work in information technology, microelectronics and physics. The SRA (Automated Reasoning Systems) division is active in the fields of knowledge representation, reasoning and learning. A research line is devoted to machine learning and adaptive systems. Original and innovative contributions have been made in budgeted learning, hierarchical classification and preference elicitation. The results of the research have been applied in the domain of document organization, information filtering and experiment design both in biology and agriculture.web-page: http://sra.itc.it/
