ABSTRACT
The first emulation of spiking LIF behavior in ultrafast optical components was a large fiber-based system that took up the area of an optical bench. Although this rudimentary photonic neuron was bulky, power hungry (2W), and inefficient (1%), it demonstrated a variety of important hybrid spike processing functions, including integration and thresholding [1–3]. Many of the same physical principles pioneered in these initial prototypes are relevant for more recently developed integrated models (Chapter 6) because of their ability to incorporate into larger neural networks (Section 10.3).
