Edge computing with optical neural networks via WDM weight broadcasting

Abstract

We introduce an optical neural-network architecture for edge computing that takes advantage of wavelength multiplexing, high-bandwidth modulation, and integration detection. Our protocol consists of a server and a client, which divide the task of neural-network inference into two steps: (1) a difficult step of optical weight distribution, performed at the server and (2) an easy step of modulation and integration detection, performed at the edge device. This arrangement allows for large-scale neural networks to be run on low-power edge devices accessible by an optical link. We perform simulations to estimate the speed and energy limits of this scheme.

Massively parallel universal linear transformations using a wavelength-multiplexed diffractive optical network

“…Also the input and output fields of each one of these linear transformations in our framework are spatially encoded in 2D at the input/output FOVs using the same wavelength, rather than being spectrally encoded, as demonstrated in earlier WDM-based designs. [71][72][73] This unique feature allows our diffractive network to all-optically perform a large group of independent linear transformations in parallel by sharing the same 2D input/output FOVs.…”

“…Moreover, we would like to clarify that the wavelength multiplexing scheme used for our framework in this paper should not be confused with other efforts that integrated wavelength-division multiplexing (WDM) technologies to optical neural computing, such as in Refs. 71–73. In these earlier work, WDM was utilized to encode the 1D input/output information to perform a vector–matrix multiplication operation, where the optical network was designed to perform only one linear transformation based on a single input data vector, producing a single output vector that is spectrally encoded.…”

Multiplexing methods for scaling up photonic logic

“…Recently, we proposed Netcast, an optical serverclient protocol for edge computing based on wavelength-division multiplexing (WDM), difference detection and integration, and optical weight delivery. 21,22 This protocol splits the computation into two components: a WDM modulator array constituting the "weight server" (Fig. 4(a)), connected via an optical link to a SWaPconstrained client (Fig.…”

Delocalized photonic deep learning on the internet’s edge

“…To address these problems, we introduce here a photonic edge computing architecture, named Netcast, to minimize the energy and latency of large linear algebra operations such as general matrix-vector multiplication (GEMV) ( 5 ). In the Netcast architecture, cloud servers stream DNN weight data ( W ) to edge devices in an analog format for ultraefficient optical GEMV that eliminates all local weight memory access ( 14 ).…”