Rough Sets and Neuroscience

Abstract

It has been almost exactly 10 years since the publication of the Neurocomputing Special Volume on Rough-Neuro Computing, and nearly 8 years since the seminal book "Rough-Neural Computing" came out. Rough-Neuro (or Neural) Computing (RNC) generalizes traditional artificial neural networks by incorporating the concepts of information granularity and computing with words. It provides solid theoretical foundations for hybridization of neural computing with the theory of rough sets, as well as rough mereology, and has many interesting practical applications. Interestingly, while the RNC paradigms directly or indirectly draw extensively from the field of neuroscience, not many applications of the theory of rough sets (in the form of RNC or otherwise) to solve problems in that field exist. This is somewhat surprising as many problems in neuroscience are inherently vague and/or ill-defined and could potentially significantly benefit from the rough sets' ability to deal with imprecise data, and those applications that have been proposed, have been very successful. In this chapter, we describe a few examples of the existing applications of the theory of rough sets (and its hybridizations) in the field of neuroscience and its clinical "sister," neurology. We also provide a discussion of other potential applications of rough sets in those areas. Finally, we speculate on how the new insights into the field of neuroscience derived with the help of rough sets may help improve RNC, thus closing the loop between the two fields. © Springer-Verlag Berlin Heidelberg 2013.