"The way information is represented, stored, recalled and processed in the neocortex is assuredly one of the most puzzling enigmas that science will have to solve during this century. Setting the basic principles of the Mental Information Theory is a high challenge potentially opening huge fields of research and progress in various domains.This research project is at the crossroads of neuroscience, computational intelligence and information theory, with a particular emphasis on the properties of distributed information processing architectures. Precisely, this project aims to identify clearly, deepen and exploit the strong analogies (distributed structure and message passing, storage capacity, discrimination ability, resilience, importance of cycles and correlation, etc.) that can be found between the structures and properties of the cerebral cortex and those of modern error correcting decoders studied in the communications science area.It is possible today to deduct from all the observations and discoveries made recently about cerebral biology a minimum material that can help information theory (communication, coding, graphs, etc.) contribute to the understanding and imitation of the neocortex functioning. In particular, the recently introduced biological concepts of neural clusters, neural cliques and sparse coding are exploited in order to devise original and efficient brain-inspired networks. We have already demonstrated that combining these concepts in a judicious approach opens the way to store and retrieve a number of messages proportional to the square of the number of neurons (to be compared for instance with the well-known sub-linear law of Hopfield networks).The objective of this project is twofold: 1) implementing electronic machines having the ability to learn a lot of information and to produce new one by association, fusion or crossbreeding within a 5-year period, 2) contributing to the understanding of the biological long and short term memories"