Reconfigurable optical neural network based on phase change material and application thereof

Abstract

The invention discloses a reconfigurable optical neural network based on a phase change material and application thereof, and belongs to the field of optical neural networks, and the reconfigurable optical neural network comprises an optical information input layer, a reconfigurable diffraction layer and a detection output layer. The reconfigurable diffraction layer comprises n optical diffraction plates which are parallel to one another and are arranged at intervals, and is used for regulating and controlling the intensity distribution of incident light to obtain plane light to be measured; the optical diffraction plate comprises a substrate, a phase-change material layer and a protective layer which are sequentially arranged, the phase-change material layer and the protective layer are divided into a plurality of independent areas to form a plurality of units, and phase-change materials in the units can be switched among a plurality of states including a crystalline state and an amorphous state; the optical information input layer is used for generating incident plane light which carries input image information and is normally incident to the reconfigurable diffraction layer; and the detection output layer is used for detecting the power at m preset positions in the to-be-detected planar light so as to determine an image classification result. According to the invention, the reconfigurable all-optical neural network can be provided, and the flexibility of the optical neural network is improved.

Description

technical field [0001] The invention belongs to the field of optical neural network, and more specifically relates to a reconfigurable optical neural network based on phase change material and its application. Background technique [0002] With the development of computer science, neural networks have been widely used in fields such as big data and image recognition in daily life. However, the serial operation mechanism adopted by modern computers based on von Neumann structure is inefficient and energy-intensive when performing neural network training and calculation based on parallel structure. Therefore, in the face of increasingly complex network structures and data The computing power and power consumption gradually entered the bottleneck. To solve this problem, all-optical neural network is considered as a potential solution. [0003] The all-optical neural network is a hardware network based on optical devices. Its advantage is that since the information is carried ...

AI Technical Summary

Problems solved by technology

At present, there are two main technical routes for all-optical neural networks. One is an on-chip solution based on the integration of waveguide structures and adjustable optical devices. This solution is highly integrated, but the network complexity and total number of devices increase rapidly with the increase of the total number of neurons. limited scalability

[0004] Another solution is based on free-space optical diffraction, which is slightly larger in size but better in scalability and reliability than the on-chip solution. However, the optical neural network realized by the proposed free-space diffraction solution is easy to design after it is manufactured. It is fixed, and the phase, amplitude, and transmission / reflection coefficient of each neuron cannot be changed, that is to say, the network weight and activation function cannot be changed. Therefore, only a single preset function can be realized. When the function changes , need to redesign a new optical neural network, less flexibility

Images