Figure 8. Pragmatic stability of network function.

Practical considerations such as fabrication defects, input angle deviation and lower amplification level may affect the function of the network. In all three examinations we compared the output of the altered device with the ideal one, using the results of pattern identification (as shown in Fig. 2). (a) Generating a stress-like warping at random angle (in the xy-plane) with random global changes in core and cladding indices did not have a major effect on the relative output reading (output 1 vs. 3 vs.5 see arrow marking) even when deformation was taken beyond what is assumed as a standard fabrication capability (1–2% deformation per meter). (b) While all previous simulation assumed that input sources re all parallel to Z-axis (ϴ = 0°), we show that mild tilting of sources (up to ϴ = 8°) still enable adequate relative outputs (all weaker but ratios are maintained). (c) We previously assumed that given weaker and more realistic amplification power, we can design a longer device with larger pitch that will retain the same function. Here we demonstrate such a device that is 12 times longer, with 20 times weaker amplification, and 14 μm pitch instead of 9 μm. Again, while overall output is weaker, the ratio between output cores is maintained (output 1 ≫ output 3 or output 5).