Apparatus for constructing an optical computing network based on photonic leads
By using a photonic lead-based construction method and employing multi-aperture jumper frames and hierarchical couplers, the nonlinearity problem of optical computing networks was solved, improving flexibility and accuracy, reducing energy consumption, and achieving efficient optical computing and image recognition.
Patent Information
- Application Number
- CN202310056863.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-01-13
- Publication Date
- 2025-12-23
- Estimated Expiration
- 2043-01-13
Smart Images

Figure CN116011539B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of optical computing network, and particularly provides an apparatus for constructing an optical computing network based on photon leads. BACKGROUND
[0002] Optical computing technology is a technology and device for realizing operation processing and data transmission by using optical methods. Optical computing has advantages of two-dimensional parallel processing, high speed, large capacity, space transmission and anti-electromagnetic interference. The main features of optical neural networks are group parallelism, high interconnection density, association and fault tolerance. The main research content is to realize a neural network model by using optical methods.
[0003] With the development and in-depth research of optical computing and neural networks, optical neural networks combine optoelectronic technology and traditional neural network models, and are expected to break through technical bottlenecks such as long delay and high power consumption of traditional electronic neural networks, and can also be used to construct optical processors. Artificial intelligence is empowered by photons, and photons or optical paths are used to replace traditional electronic computing to realize more efficient artificial intelligence computing.
[0004] Compared with the currently more mature electronic neural network, the photon neural network still has broad space for improvement in trainability, integration, scaling and practicality. The non-ideal performance and low stability of optoelectronic devices inhibit the trainability, integration and scaling of photon neural networks, and have strict requirements for constructing neural network models with more complex functions. Unlike electronic neural networks, the application field of photon neural networks is also limited by all-optical nonlinearity.
[0005] In addition to the influence of the nonlinearity problem, the current apparatus for constructing an optical computing network also has problems of high equipment cost, low flexibility, poor adjustability, etc. In addition, the existing equipment often uses diffraction methods such as diffraction gratings and spatial light modulators for optical information recognition and calculation. However, in the photon neural network, multi-stage diffraction is used, which has problems of low diffraction efficiency and serious inter-stage crosstalk. In addition, each time the network is adjusted, the diffraction grating of each stage needs to be adjusted, which is extremely tedious and has low precision. SUMMARY
[0006] To solve the above problems, the present application provides an apparatus for constructing an optical computing network based on photon leads, which mainly realizes different coupling efficiencies by controlling the position of the single-mode fiber jumper head, realizes nonlinear data extraction through the convolution characteristics of the photon coupling efficiency, and realizes flexible network adjustment by using plug-in single-mode fiber jumper heads. By using hierarchical coupling instead of multi-stage diffraction, the problems of low diffraction efficiency and serious inter-stage crosstalk are avoided.
[0007] The application provides a device for constructing an optical computing network based on a photon lead, comprising a multi-aperture jumper rack, a plurality of single-mode fiber jumper heads, a nonlinear modulator and an interference device.
[0008] Photons are injected into the multi-aperture jumper rack, and the multi-aperture jumper rack performs wavefront mass distribution on the photons to reduce errors in optical computing.
[0009] A plurality of aperture interfaces are formed on the multi-aperture jumper rack, the single-mode fiber jumper heads are connected to the aperture interfaces in a plug-in manner, the rear end of the single-mode fiber jumper head is connected to an optical fiber for transmitting optical signals to the nonlinear modulator, a plurality of couplers are arranged in the nonlinear modulator, the couplers are arranged in a hierarchical manner, and the couplers couple the photons in the optical fiber; the single-mode fiber jumper head has a six-dimensional adjustment function, the coupling efficiency of the photons is adjusted by adjusting the position of the single-mode fiber jumper head, and nonlinear data extraction is realized through the convolution characteristics of the coupling efficiency of the photons.
[0010] The multi-channel photons subjected to nonlinear processing are subjected to interference by the interference device, and finally a plurality of point light intensities are emitted.
[0011] Preferably, the photons are injected into the multi-aperture jumper rack through a self-collimation system to ensure the parallelism of the photons.
[0012] Preferably, the single-mode fiber jumper head is internally provided with a six-dimensional adjuster for realizing six-degree-of-freedom precise adjustment of the single-mode fiber jumper head.
[0013] Preferably, the six-dimensional adjuster adjusts the position of the single-mode fiber jumper head according to the coincidence degree of a theoretical mode field of the optical fiber and an energy distribution mode field of the input photons, and the higher the coincidence degree of the theoretical mode field of the optical fiber and the energy distribution mode field of the input photons, the higher the coupling efficiency of the photons.
[0014] Preferably, the interference device adopts an interferometer.
[0015] Preferably, a point detector is used to receive the output point light intensity to obtain the calculation result of the device for constructing an optical computing network based on a photon lead.
[0016] Preferably, the rear end of the point detector is connected to an electrical network or an energy judgment circuit, and the image content of the photon input end is judged according to the calculation result output by the device for constructing an optical computing network based on a photon lead.
[0017] Preferably, all the optical fibers can adopt polarization maintaining optical fibers, or a part of the optical fibers adopt polarization maintaining optical fibers, and a polarization controller is arranged on the optical fiber to realize separate measurement of different polarization states and improve the information flux of detection.
[0018] Preferably, the coupler adopts a hierarchical mode, specifically: the first-stage coupler groups and couples multiple photons, the lower-stage coupler groups and couples the coupled light obtained by the upper-stage coupling, and the last-stage coupler completes the coupling.
[0019] Compared with the prior art, the application can achieve the following beneficial effects:
[0020] The application provides a device for constructing an optical computing network based on a photon lead, which can completely replace an electrical network to perform data processing and computing, so that the operation speed of the network is improved and the overall energy consumption of the network is reduced.
[0021] The application improves the flexibility and adjustability of the network through the design of the pluggable single-mode optical fiber jumper head, has high precision, and adopts hierarchical coupling instead of multi-stage diffraction to avoid problems such as low diffraction efficiency and serious inter-stage crosstalk, because the coupler replaces the diffraction grating, and the economic cost is greatly reduced.
[0022] The application can realize image recognition, energy calculation, information processing and other functions by connecting different circuits or devices, and has good application prospects in many fields. DETAILED DESCRIPTION
[0023] Figure 1 is a structure diagram of the device for constructing an optical computing network based on a photon lead provided by the application.
[0024] The reference signs in the drawings include:
[0025] Input end photon 1, multi-aperture jumper rack 2, single-mode optical fiber jumper head 3, optical fiber theoretical mode field distribution diagram and energy distribution mode field distribution diagram of the input photon 31, optical fiber 4, nonlinear modulator 5, interference device 6. DETAILED DESCRIPTION
[0026] Hereinafter, embodiments of the application will be described with reference to the accompanying drawings. In the following description, the same modules are denoted by the same reference signs. In the case of the same reference signs, their names and functions are also the same. Therefore, the detailed description thereof will not be repeated.
[0027] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and do not constitute a limitation on the application.
[0028] Figure 1 A simplified structure of the device for constructing an optical computing network based on a photon lead provided by the application is shown.
[0029] As Figure 1As shown, the device for constructing an optical computing network based on a photon lead wire provided by the application mainly comprises: a multi-aperture jumper rack 2, a single-mode optical fiber jumper head 3, an optical fiber 4, a nonlinear modulator 5 and an interference device 6.
[0030] The input photon 1 can be a transmission light source of a single-mode optical fiber or an emission light source such as a laser. The input photon 1 is incident on the multi-aperture jumper rack 2, and a self-collimation system is constructed between the input photon 1 and the multi-aperture jumper rack 2 to correct the input photon 1 into a light beam with good laying parallelism. The multi-aperture jumper rack 2 performs wavefront quality distribution on the photon to ensure that the point spread function introduced by the wavefront does not affect the detection error of the device, i.e., the optical computing error, thereby avoiding the generation of a bit error rate. The multi-aperture jumper rack 2 is provided with a plurality of aperture interfaces. Different aperture interfaces correspond to photons received at different positions on the back of the multi-aperture jumper rack 2. The single-mode optical fiber jumper head 3 is connected to the aperture interfaces in a pluggable manner. When only part of the photon information at different positions needs to be detected, the single-mode optical fiber jumper head 3 only needs to be inserted into the aperture interface at the corresponding position. The pluggable manner can flexibly realize the change of the optical network and avoid complex optical network calibration. The rear end of the single-mode optical fiber jumper head 3 is connected to the optical fiber 4. The optical fiber 4 can be a 150 single-mode optical fiber. The optical fiber 4 transmits the received photons to the nonlinear modulator 5. The optical fiber 4 and the nonlinear modulator 5 are also connected in a pluggable manner. The optical network can also be changed at this position. When part of the optical fiber is not working, it can be removed and saved. The nonlinear modulator 5 mainly comprises a plurality of couplers. The couplers can be a plurality of high-efficiency couplers. The couplers are arranged in a hierarchical manner. The hierarchical manner is that the photons first enter the first-stage coupler. The first-stage coupler groups and couples two or more photons. Each coupler corresponds to the same number of coupled light paths. The lower-stage coupler groups and couples the coupled light obtained by the upper-stage coupler. The last-stage coupler completes the coupling. The number of the last-stage couplers can be designed according to requirements.
[0031] The single-mode optical fiber jumper head 3 is also provided with a six-dimensional adjuster for realizing the six-degree-of-freedom precise adjustment of the single-mode optical fiber jumper head, i.e., the precise adjustment of the specific position of the single-mode optical fiber jumper head 3 on the multi-aperture jumper rack 2.
[0032] The six-dimensional adjuster adjusts the position of the single-mode fiber jumper head 3 according to the coincidence degree of the fiber theory mode field and the energy distribution mode field of the input photons, as shown in the fiber theory mode field distribution graph and the input photon energy distribution mode field distribution graph 31. The change of the position of the single-mode fiber jumper head 3 will affect the coincidence degree of the fiber theory mode field and the energy distribution mode field of the input photons. The higher the coincidence degree, the higher the coupling efficiency of the photons. The lower the coincidence degree, the lower the coupling efficiency of the photons. That is, the coupling efficiency of the photons can be adjusted by adjusting the position of the single-mode fiber jumper head 3, and the linear to nonlinear transformation can be realized through the convolution characteristics of the photon coupling efficiency, and the nonlinear data obtained can ensure the normal operation of the optical network.
[0033] The multi-channel photons processed by the nonlinear processor are transmitted to the interference device 6 through the fiber 4 for interference. The interference device 6 adopts the existing Michelson interferometer. After interference, a plurality of point light intensities are emitted. The plurality of point light intensities are the calculation results of the optical network, that is, carry the information obtained by the optical network calculation. The point detector is used to receive the point light intensity, and the calculation result of the device for constructing an optical computing network based on photon leads is read out. The back end of the point detector can be connected to an electrical network or an energy judgment circuit. According to the calculation result output by the device for constructing an optical computing network based on photon leads, the image content of the photon input end is judged. In addition, the subsequent network can also not be connected. The intensity of each point light intensity reflects the characteristic information of the input image. The image content of the input end can be judged through the characteristic information. However, this method is relatively tedious and time-consuming. It is more convenient to judge the image content through the connection of the electrical network or the energy judgment circuit. The electrical network or the energy judgment circuit connected subsequently belongs to the prior art, and will not be described here.
[0034] The training process of the device of the application to the optical network belongs to the prior art, and can be trained according to the existing mature electrical network training process. Here, it will not be described here.
[0035] In addition, all the fibers 4 can be polarization maintaining fibers, or a part of the fibers 4 can be polarization maintaining fibers. One or a plurality of polarization controllers are arranged on the fibers 4. The device for constructing an optical computing network based on photon leads can realize the measurement of different polarization states respectively, improve the detection information flux, and improve the image recognition accuracy through the measurement of different polarization states respectively.
[0036] The application can completely replace the electric network to process data and calculate, so that the operation speed of the network is improved and the overall energy consumption of the network is reduced, and meanwhile, many problems in the existing optical network are solved, the application has higher flexibility and calculation precision, meanwhile, hierarchical coupling is adopted instead of multi-stage diffraction, so that the problems of low diffraction efficiency and serious inter-stage crosstalk are avoided, and the high economic cost caused by a large number of diffraction gratings is also avoided. The application can realize image recognition, energy calculation, information processing and many other functions by connecting different circuits or devices, and has good application prospect in many fields.
[0037] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.
[0038] The specific embodiments of the application described above do not constitute a limitation on the scope of protection of the application. Any various other corresponding changes and modifications made according to the technical concept of the application shall be included in the scope of protection of the claims of the application.
Claims
1. A device for constructing an optical computing network based on photonic leads, characterized in that, The application relates to a multi-aperture jumper rack, a plurality of single-mode optical fiber jumper heads, a nonlinear modulator and an interference device. Photons are injected into the multi-aperture jumper rack, and the multi-aperture jumper rack performs wavefront mass distribution on the photons to reduce errors of optical calculation. A plurality of aperture interfaces are formed on the multi-aperture jumper rack, the single-mode optical fiber jumper heads are connected to the aperture interfaces in a plug-in manner, the rear end of the single-mode optical fiber jumper head is connected with an optical fiber, and the optical fiber is used for transmitting an optical signal to the nonlinear modulator; a plurality of couplers are arranged in the nonlinear modulator, the couplers are arranged in a hierarchical manner, and the couplers couple photons in the optical fiber; the single-mode optical fiber jumper head has six-dimensional adjustment functions, the coupling efficiency of the photons is adjusted by adjusting the position of the single-mode optical fiber jumper head, and nonlinear data extraction is realized through convolution characteristics of the coupling efficiency of the photons. After being subjected to nonlinear processing, the multi-path photons are subjected to interference through the interference device, and finally a plurality of point light intensities are emitted. The photons are injected into the multi-aperture jumper rack through a self-collimation system, so as to ensure the parallelism of the photons.
2. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 1, wherein, The single-mode optical fiber jumper head is internally provided with a six-dimensional adjuster, so as to realize six-degree-of-freedom precise adjustment of the single-mode optical fiber jumper head.
3. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 1, wherein, The six-dimensional adjuster adjusts the position of the single-mode optical fiber jumper head according to the coincidence degree of a theoretical mode field of the optical fiber and an energy distribution mode field of the input photons, and the higher the coincidence degree of the theoretical mode field of the optical fiber and the energy distribution mode field of the input photons, the higher the coupling efficiency of the photons.
4. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 3, wherein, The interference device adopts an interferometer.
5. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 1, wherein, A point detector is used to receive the output point light intensity, and the calculation result of the device for constructing an optical calculation network based on photon leads is obtained.
6. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 1, wherein, The rear end of the point detector is connected with an electric network or an energy judgment circuit, and the image content of the photon input end is judged according to the calculation result output by the device for constructing an optical calculation network based on photon leads.
7. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 6, wherein, The optical fibers can all be polarization maintaining optical fibers, or a part of the optical fibers are polarization maintaining optical fibers, and a polarization controller is arranged on the optical fibers, so as to realize separate measurement of different polarization states and improve the detection information flux.
8. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 1, wherein, The couplers are arranged in a hierarchical manner, that is, the first-stage couplers group and couple the multi-path photons, the lower-stage couplers group and couple the coupled light obtained by the upper-stage coupling, and the last-stage couplers complete the coupling.
9. The apparatus for constructing an optical computing network based on photonic pig-tailing of claim 1, wherein,