Analog Neural Networks

Neuronal schemes realizing operations with continuous quantities are presented. These schemes are based on an analog neuron as a "diode-like" summator of continuous quantities (spike frequencies). Analog (continuous) logic, non-linear feedback, and neuronal structures, which can realize the complex features of information filtration, are discussed. Special attention is paid to factorial switches and synthesis of neuronal structures. Neuronal structures processing n-dimensional continuous vectors by non-linear feedbacks can realize the factorial switch, which stores and reproduces information about decreasing order of the components of the vectors.