High precision analog optical computing method and system based on bit slice

Abstract

A high-precision analog optical computing method and system based on bit slicing comprises decomposing high-precision signals into low-precision signal segments, decomposing high-precision weights into multiple low-precision weight segments, performing low-precision signal and weight multiplication operations in the optical domain respectively, and finally combining these low-precision calculation results to restore high-precision output. By splitting high-precision data into multiple low-precision bit slices for processing, it takes advantage of high-speed parallel characteristics of photonic computing to simultaneously handle multiple input bit slices and weight bit slices, improves computational efficiency of large-scale computing tasks and reduces the hardware requirements for optical computing. By adjusting values of M and N, it flexibly adapts to computing tasks of different scales and precisions, demonstrating strong adaptability and flexibility. It achieves balance between computational accuracy and hardware complexity, and fully exploits parallel processing advantages of photons, providing solutions for large-scale data processing and efficient optical domain computing.