Contatto di riferimento: Prof. Stefano Diciotti
Abstract
The first part of the seminar will introduce linear classifiers and, in particular, support-vector machines (SVMs). SVMs are state-of-the-art efficient and robust classifiers able to perform linear and non-linear classification. They have been used in a wide variety of applications including OCR, text categorization, image classification, biomedical research and diagnosis.
The second part of the seminar will present an example of the use of SVMs in the field of healthcare. In particular, it will describe a prediction model based on SVM classifiers and Generalized Linear Models which won the 2012 PhysioNet/CinC Challenge about patient-specific prediction of in-hospital mortality using general descriptors and measurements collected during the first 48 hours in the intensive-care unit.
About the speaker
Luca Citi received a degree in Electronic Engineering in 2004 from the University of Florence (Italy). His master’s thesis was about an ERP-based brain-computer interface. In 2009, he obtained a PhD from Scuola Superiore Sant’Anna and IMT Lucca (Italy) with a thesis about the decoding of neural signals for the control of robotic arm prostheses. He then spent three years as post-doc at Harvard Medical School and Massachusetts Institute of Technology (USA) working on statistical analysis of point processes (stochastic processes representing discrete events in time) applied to heartbeat series and neural spike trains. He is currently a Lecturer in Computational Intelligence at the University of Essex (UK).