Realization of Optical DEMUX using Machine Learning Model: A Future Paradigm for Optical-VLSI System

Abstract

ABSTRACT Machine learning using the ANN approach is presented in this paper to design 101 channels of optical DEMUX in the range of 1300–1400 nm. This work is used to envisage the optical parameters of silicon grating structure such as the thickness of SiO and Si which are the decisive factors to allow the desired signal through the structure to realize optical DEMUX. This is accomplished by training an ANN model with a few pre-defined optical DEMUX parameters which are obtained beforehand by the plane wave expansion method. Here, the ANN model is implemented in the Tensor Flow framework of Google Colaboratory Network (Colab) which consists of five hidden layers of ReLU activation besides the input and output linear activation nodes. Besides this, the optical parameters collected from the ANN model are used to simulate through the PWE method for validation by observing the reflectance characteristics curve and found that most of the values are satisfied to realize the optical VLSI DEMUX. The outcomes of the present work could be beneficial in the field of optical VLSI systems pertaining to the optical interconnect.