Neuromorphic Computing Based on Wavelength-Division Multiplexing
Recent advances in WDM-based ONNs are reviewed, focusing on methods that use integrated microcombs to implement ONN's, and results for human image processing using an optical convolution accelerator operating at 11 Tera operations per second are presented.
Abstract
Optical neural networks (ONNs), or optical neuromorphic hardware\naccelerators, have the potential to dramatically enhance the computing power\nand energy efficiency of mainstream electronic processors, due to their\nultralarge bandwidths of up to 10s of terahertz together with their analog\narchitecture that avoids the need for reading and writing data back and forth.\nDifferent multiplexing techniques have been employed to demonstrate ONNs,\namongst which wavelength division multiplexing (WDM) techniques make sufficient\nuse of the unique advantages of optics in terms of broad bandwidths. Here, we\nreview recent advances in WDM based ONNs, focusing on methods that use\nintegrated microcombs to implement ONNs. We present results for human image\nprocessing using an optical convolution accelerator operating at 11 Tera\noperations per second. The open challenges and limitations of ONNs that need to\nbe addressed for future applications are also discussed.\n