Hard- and software emulation of dynamic neural networks

The authors propose a hardware and software neuron-like system (APNS). The APNS system includes a neuroprocessor, a structured memory, a memory controller, an afferent signal processor, and a modification processor (IBM PC/AT). The processing of data is performed by the system in a parallel-sequential regime; the data processed are represented as vectorized massives. The authors discuss the macroarchitecture adequate to the problem, and describe its application to the realization of different neurodynamic phenomena and neuroparadigms of associative memory and learning