Some Brain Mathematics

Claude Berrou, a professor at the Ecole Nationale des Télécommunications, of Brittany, and a member of the French Academy of Sciences, developed 'turbo codes' (error-correcting codes used in mobile phones). His research on information processing and coding incited him to take a fresh look at the workings of the human brain. Vincent Gripon's research interests are graph theory, information theory, formal neural networks and new forms of artificial intelligence. He has a doctorate in information and communication sciences and technology. We know practically everything about neurons - and almost nothing about how they process mental information. Under what 'material' form does the human brain store information? How is that information internally organised and how is it retrieved on demand? Such questions concerning mental information would seem to stem from information technology, originally developed by telecommunications and coding experts, rather than from biology and neuro-anatomy. This highly accessible book offers the first mathematically coherent and biologically plausible explanation of how neural networks affix and retrieve items of knowledge. The result is an original theory, combining neurons and graphs, error-correcting codes and cortical columns, neural 'clicks' and other exploits, in the on-going search for the algorithms of the human brain. The theory - and the entirely digital model of cerebral memory that it leads to - offers numerous promising developments, in the neurosciences and the domain of artificial intelligence alike.