Experimental Simulation for Two Optically Filtered Modulation Weights in Laser Diode as a Self-Learning Layer

Abstract

In this study, the response of a nonlinear laser medium is experimentally studied. In the study, a hybrid version of the input layer that multiplies optically and accumulates electrically is compared with a wholly optical version that multiplies and accumulates optically. This medium is subjected to two different paths of optically filtered and attenuated feedback. With such a system, the variation of feedback weight in one of them is tested in correspondence to the second one. Observations for frequency spectra are carried out to simulate the resultant response with an input layer for a neural network based on chaotic carriers. Chaotic laser emission was observed as a function of several control parameters, which are D.C. bias voltage, branch optical attenuation, and feedback strengths based on filtration with fiber Bragg grating. This learning rule is linear in the difference between each input and output of a neuron. This is an enhancing/inhibiting rule. The thresholds are adjusted in such a way that the output of the neuron is either pushed in the same direction as the input (enhancing) or pushed in the opposite direction (inhibiting).

Keywords

• Machine Learning
• Neural Network
• Chaotic Modulation
• Laser Diode