A neural processing unit for self-organizing maps

International audienceFace to the limitations of the classical computationmodel, neuromorphic systems are envisaged as an alternative.Interesting properties of the neural systems are its highly parallelcomputation, and its self-organizing capabilities. In this paper,we study how to bring distribution in the hardware computationof these neural models, especially in the case of self-organizingmaps. We propose a bio-inspired hardware substrate in which agrid of neural processing elements will support a set of neurocognitiveprocesses. We describe an original distributed artificialneural network adapted to hardware scalability and provide theresults of its implementation as a set of NPUs onto FPGA devices