The bifurcating neuron: building block for a new generation of artificial neural networks

Summary form only given. The bifurcating neuron concept offers potential through: preservation of the time-dimension; exhibiting a wide-range of firing modalities similar to those observed in physiological studies of the living neuron and can bifurcate between these firing modalities depending on the nature of its activation, behaving thereby as a complex encoder of the spatiotemporal activity of its synaptic inputs; offering structural simplicity and therefore readily constructed. These attributes of the bifurcating neuron are discussed using one simple possible optoelectronic realization. It is shown how it exhibits, depending on the nature of its input, phase-locked periodic firing, quasi-periodic firing, bursting, erratic or chaotic firing, and extinction in addition to conventional sigmoidal dependence of firing frequency on activation potential