Computer Science Department
UniversitÓ degli studi di Milano
Via Comelico 39
Tel: (+39) 02 503 16295/16321
(replace - at - with @ )
research interests of Marco Frasca involve the design and development
of machine learning methodologies for modelling relevant problems in
Molecular Biology and Medicine.|
A particular interest has been shown for the design of Hopfield based models to address different issues raising in the context of automatic inference. A family of parametric Hopfield networks has been developed for dealing with classification problems with high diproportion between positive and negative labels, and a novel class of Hopfield networks has been studied, namely the "multi-category Hofield network", to exploit prior property-driven partitions of input instances.
Recently, he also designed network-based methodologies for determining genes responsible for human genetic disorders.
Although mostly conceived in the framework of computational biology problems, the developed algorithms are general enough to be applied in contexts where inductive and transductive data analysis can play a key role.