Modified engagement method for matrix operation

Abstract

We describe a modified engagement method for matrix operation based on a two-dimensional crossed-ring interconnection network. Our method incorporates fewer steps than that reported by Bocker et al. 3Appl. Opt. 22, 804 1198324, and its performance is found to be the most efficient 1minimum steps2 in comparison with other systolic and@or engagement methods for matrix operation. Thus, it may be helpful for other optical and electronic implementations of matrix operations. One compact optoelectronic integrity approach for implementing the modified engagement method is briefly described.r 1995 Optical Society of America Matrix operation is one of the basic operations in scientific computing, so how to calculate it effectively is always an interesting subject. In electronics, there are a number of non-von Neumann architectures proposed for parallel matrix operations. 1 A systolic array with fixed, regular, and local interconnections is one of the prominent practical solutions that can be used efficiently for parallel matrix operation because of advanced VLSI technology. In optics, more freedom with interconnections is possible.

Optical multi-imaging–casting accelerator for fully parallel universal convolution computing

“…It is also possible to load multiple kernels on the rst SLM, and thus the convolutions between multiple kernels and multiple input channel feature maps could be realized in parallel by lling an appropriate number of zero elements between any two adjacent kernels. Furthermore, it is worth noting that considering the actual hardware scale, it is often necessary to split and reorganize the input feature map to further improve the hardware utilization, that is, to load different matrix combinations to the SLMs to execute the convolution process 50 .…”