Figure 9. Experimental validation of our conceptual “neuron-like” unit.

(a) We implemented one in-fiber neuron unit which comprises two separately amplified input cores and an integration core using single mode fibers, Y-couplers and two EDFAs. (b) We chose low power inputs so that the amplifiers will function within their linear amplification region. EDFA outputs without inputs (Out₀) were subtracted from all outputs with inputs to examine the effect of inputs (ED = EDFA). The y-axis variance stems from the different levels of amplification. (c) The “device” output shows linear dependency on EDFA2 inputs while EDFA1 inputs contributed mostly to the y-axis variance. It is clear that while stronger amplification results in greater influence of the inputs (cyan graph), linearity in weak amplification becomes sub-linear with stronger amplification. (d) Combining two levels of inputs and two levels of amplification into a color-mapped truth table demonstrates that both input level and amplification level determine the output of the modeled unit.