Optical Implementation Of Matrix-Vector And Matrix-Matrix Multiplication

Abstract

From the viewpoint of information processing, the main advantage of an optical processor is its ability to operate in a highly parallel way. One of the most important applications of an optical processor, where its parallelism is efficiently used, is considered to be linear transformation because it involves many multiplications and additions in parallel. Matrix multiplication is a fundamental operation of such a linear transformation.

In the present paper, a simple matrix multiplication scheme in binary number is proposed and demonstrated. In this case the input 1 and 2, i. e, the elements of two matrices are to be coded using the coding scheme as proposed by Tanida et al. Though the matrices to be multiplied are binary i.e, its elements can have two states ‘one’ and ‘zero’, the output matrix is not a binary one; hence, the decoding system will be different. Some experimental results for matrix-vector and matrix-matrix multiplication are also presented.