Structural and Schematic Methods of the Biological Neuron Networks Properties Projecting on the Architecture of Modern Integrated Neuron-Like Systems with Analog-Digital Information Processing

This paper substantiates the essential role of spacetime dependence of the information representation form in the development of intellectual energy-efficient and noise-resistant data processing systems of large dimensions. The authors propose the artificial neuron networks design approach with local analog and global digital neural data processing (LAGDNDP). According to LAGDNDP approach, we interpret the role of different forms of neural networks plasticity as the basis of global and local network adaptation to external conditions of functioning, internal parameters variation, network elements failure and degradation. We show the primary role of structural and circuit design methods in the development of highly integrated neuron-like architectures of large dimensions with advanced scalability, local and global adaptation. LAGDNDP approach together with modern integrated analog and digital technologies supports structural self-organization while training on the input data flows, the formation of a hierarchical fractal structure, emulation of multi-level spatialtemporal dynamics, sub-threshold information processing.