Hybrid integrated photonic pulse neural network device and digital image classification method

By designing a hybrid integrated photonic pulse neural network device, and utilizing MZI networks and VCSEL-SA to realize the photonic pulse neural network, the problem of achieving it using purely optical methods in the existing technology has been solved. This enables efficient, low-power photonic pulse neural network computation and large-scale network expansion, making it suitable for tasks such as digital image classification.

CN115393602BActive Publication Date: 2026-07-03XIDIAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
XIDIAN UNIV
Filing Date
2022-07-12
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

There is currently no feasible solution to realize photonic spiking neural networks using purely optical means, and there is a lack of methods for realizing large-scale photonic spiking neural networks.

Method used

A hybrid integrated photonic spiking neural network device was designed, including an optical encoding module, an optical synapse module, and an optical response output module. The neuron function is realized by using an MZI network and VCSEL-SA. The weight matrix is ​​adjusted by singular value decomposition and ReSuMe supervised learning algorithm to realize the task function of the photonic spiking neural network.

Benefits of technology

It achieves efficient and low-power photonic pulse neural network computing, enabling digital image classification, and can be extended to large-scale photonic pulse neural networks through time-division multiplexing, making it suitable for a variety of application scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115393602B_ABST
    Figure CN115393602B_ABST
Patent Text Reader

Abstract

The application discloses a hybrid integrated photon pulse neural network device, comprising: an optical coding module comprising N presynaptic neurons; an optical response output module comprising N postsynaptic neurons; the neurons are VCSEL-SAs, and N is greater than or equal to 1; an optical synapse module comprising a first MZI subnetwork, an MZI array and a second MZI subnetwork in sequence, forming an MZI network with N inputs and N outputs; the first MZI subnetwork and the second MZI subnetwork each comprise N(N-1) / 2 MZIs arranged in a Reck architecture, and the MZI array comprises N MZIs arranged in parallel; the phase parameters of the MZIs in the optical synapse module are obtained by singular value decomposition of an N*N weight matrix, construction of an equation group containing trigonometric functions based on a unitary matrix obtained by decomposition and solution of the equation group.
Need to check novelty before this filing date? Find Prior Art