Optical neural network all-optical nonlinear activation layer and implementation method thereof

Abstract

The invention discloses an optical neural network all-optical nonlinear activation layer and an implementation method thereof. The MZI waveguide configuration is combined with the graphene heterogeneous enhanced Bi2Te3 nonlinear material, and the nonlinear response of the graphene heterogeneous enhanced Bi2Te3 nonlinear material is further amplified by utilizing the on-chip waveguide structure design, so that the design of the on-chip integrated nonlinear activation layer is completed, the optical nonlinear calculation in the on-chip integrated waveguide is realized, the problem that the nonlinear degree of an optical nonlinear material is weak is solved, the function of an optical neural network is expanded, and possibility is provided for use of a multi-layer pure optical neural network; and the optical nonlinear activation layer provided by the invention not only can be used for an on-chip integrated optical neural network, but also can be used for scenes needing nonlinear calculation in other integrated optical signal processing platforms, is extremely high in response speed, and can meet the requirements of low-energy-consumption and high-speed calculation.

Description

technical field [0001] The invention relates to optical signal processing technology, in particular to a design and implementation method for an all-optical nonlinear activation layer in an on-chip optical neural network. Background technique [0002] The main structures in most existing on-chip waveguide-integrated optical neural networks and other on-chip optical computing platforms are beam-splitting waveguide units and cascaded Mach-Zehnder interferometers (MZI), which are only suitable for linear calculate. Since the superposition of linear calculation is still linear calculation, regardless of the total number of layers in the above design, its calculation result can be equivalent to a matrix multiplication operation, and the parameter range is limited, which cannot meet the needs of neural network for data fitting. Therefore, this kind of optical neural network relies on further nonlinear calculations in electronic circuits during subsequent information processing, a...

AI Technical Summary

Problems solved by technology

However, this kind of phase change material is non-volatile. If the last pulse exceeds the threshold to make it enter the amorphous state, it needs additional input energy to reset to the crystalline state, so it cannot be operated continuously, and the response time and energy consumption cannot meet high-speed computing needs

[0004] The reason why it is difficult to perform nonlinear calculations on an optical platform is that the all-optical nonlinear effect of materials is weak, and there is a lack of strong enough nonlinear materials, so it is difficult to achieve strong nonlinear effects in on-chip integrated devices; while it can be used for nonlinear GST materials for linear calculations have non-volatile properties, and are also not suitable for efficient calculations with fast responses

Images