Coding module, input port and three-dimensional input port for optical neural network

Abstract

The invention discloses a coding module for an optical neural network, an input port and a three-dimensional input port. The coding module comprises a plane wave generator; an information encoder; the plane wave generator is arranged at the upstream of the information encoder along the light path propagation direction; the plane wave generator is used for converting the received light beams into plane waves, and the information encoder is used for carrying out information encoding on the plane waves; wherein the information encoder is a reconfigurable metasurface, and the reconfigurable metasurface comprises at least two reconfigurable micro-nano units. According to the method, the light beam is converted into the plane wave before information coding, so that the precision of information coding is improved, and the accuracy of processing a light calculation task is further improved; moreover, the adjustable metasurface serves as an information encoder, so that the number of information which can be borne in a unit area is larger, the input information density of the optical neural network is improved, a data structure with higher dimensionality or more complexity can be processed, and the optical neural network can process a large-scale optical calculation task.

Description

Technical Field

[0001] The present application relates to the field of optical chips, and particularly to an encoding module, an input port, and a three-dimensional input port for an optical neural network. Background Art

[0002] In recent years, with the in-depth development of optical technologies and metamaterial research, various studies on using optical structures to simulate neural networks, namely optical neural networks, have been carried out. Optical neural networks use photons as information carriers and achieve signal transmission, modulation, and processing through optical components.

[0003] However, due to current hardware and technological limitations, such as insufficient integration of photonic chips and insufficient modulation ability of optical components for light beams, the input information density of optical neural networks is relatively low, resulting in a small amount of data that can be processed by optical neural networks per unit time and making it difficult to meet the requirements of large-scale optical computing tasks. Summary of the Invention

[0004] In view of the above technical problems, the present application provides an encoding module, an input port, and a three-dimensional input port for an optical neural network. Before information encoding, the present application converts a light beam into a plane wave, avoiding the influence of the original wavefront curvature of the light beam on the accuracy of information encoding, which may lead to a decrease in the accuracy of the final optical computing task; and, by using a tunable metasurface as the information encoder, the number of information that can be carried per unit area is increased, the input information density of the optical neural network is improved, enabling the optical neural network to process more data streams, and thus being able to process higher-dimensional or more complex data structures, so that the optical neural network can handle large-scale optical computing tasks.

[0005] According to one aspect of the embodiments of the present application, an encoding module for an optical neural network is disclosed, the encoding module including: a plane wave generator; an information encoder;

[0006] The plane wave generator is disposed upstream of the information encoder along the optical path propagation direction;

[0007] The plane wave generator is configured to convert a received light beam into a plane wave, and the information encoder is configured to perform information encoding on the plane wave;

[0008] Wherein, the information encoder is a reconfigurable metasurface, and the reconfigurable metasurface includes at least two reconfigurable micro-nano units.

[0009] In an exemplary embodiment of the present application, one reconfigurable micro-nano unit on the reconfigurable metasurface constitutes a data point.

[0010] In an exemplary embodiment of the present application, at least two reconfigurable micro-nano units on the reconfigurable metasurface constitute a data point.

[0011] In an exemplary embodiment of the present application, the plane wave generator is composed of at least one metasurface and / or at least one refractive lens.

[0012] In an exemplary embodiment of the present application, the equivalent refractive index of the reconfigurable micro-nano unit changes when subjected to an external excitation.

[0013] In an exemplary embodiment of the present application, the external excitation is one or more of optical excitation, electrical excitation, thermal excitation, and mechanical excitation.

[0014] According to one aspect of the embodiments of the present application, an input port for an optical neural network is disclosed, characterized in that the input port includes: an encoding module as described in any one of claims 1-6; an incident coupler; an output coupler;

[0015] Wherein, the incident coupler and the output coupler are connected by an optical waveguide;

[0016] The incident coupler is used to receive the light beam emitted by the light source, and the output coupler is used to emit the light beam to the encoding module.

[0017] According to one aspect of the embodiments of the present application, a three-dimensional input port for a three-dimensional optical neural network is disclosed, characterized in that the three-dimensional input port includes: at least two input ports as described in claim 8.

[0018] In an exemplary embodiment of the present application, the three-dimensional input port further includes: a diffraction element;

[0019] Wherein, the diffraction element is used to converge the light beam emitted by the light source onto the incident couplers of each of the input ports.

[0020] In an exemplary embodiment of the present application, the diffraction element is a metasurface or a diffraction grating.

[0021] The present application discloses an encoding module, an input port, and a three-dimensional input port for an optical neural network. The encoding module includes: a plane wave generator; an information encoder; the plane wave generator is arranged upstream of the information encoder along the optical path propagation direction; the plane wave generator is configured to convert the received light beam into a plane wave, and the information encoder is configured to perform information encoding on the plane wave; wherein, the information encoder is a reconfigurable metasurface, and the reconfigurable metasurface includes at least two reconfigurable micro-nano units. The present application provides an encoding module, an input port, and a three-dimensional input port for an optical neural network. Before performing information encoding, the present application converts the light beam into a plane wave, avoiding the influence of the original wavefront curvature of the light beam on the accuracy of information encoding, resulting in a decrease in the accuracy of the final optical computing task, that is, improving the accuracy of information encoding, and further improving the accuracy of processing optical computing tasks; and, by using a tunable metasurface as the information encoder, more information can be carried per unit area, improving the input information density of the optical neural network, enabling the optical neural network to process more data streams, and further enabling the processing of higher-dimensional or more complex data structures, so that the optical neural network can process large-scale optical computing tasks.

[0022] Other features and advantages of the present application will become apparent from the following detailed description, or will be learned in part through the practice of the present application.

[0023] It should be understood that the above general description and the following detailed description are only exemplary and do not limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By referring to the accompanying drawings and describing its exemplary embodiments in detail, the above and other objects, features, and advantages of the present application will become more apparent.

[0025] Figure 1 The schematic diagram of the encoding module provided by an embodiment of the present application is shown.

[0026] Figure 2 The working schematic diagram of the encoding module provided by an embodiment of the present application is shown.

[0027] Figure 3 The schematic diagram of the reconfigurable metasurface provided by an embodiment of the present application is shown.

[0028] Figure 4 The schematic diagram of the input port provided by an embodiment of the present application is shown.

[0029] Figure 5 The schematic diagram of the input port provided by an embodiment of the present application is shown.

[0030] Figure 6Shows a schematic diagram of a three-dimensional input port provided by an embodiment of the present application.

[0031] Figure 7 Shows a schematic diagram of a three-dimensional input port provided by an embodiment of the present application.

[0032] Reference numerals:

[0033] 100 - Encoding module; 101 - Plane wave generator; 102 - Information encoder; 200 - Plane wave; 300 - Input port; 301 - Incident coupler; 302 - Output coupler; 400 - Three-dimensional input port; 500 - Optical neural network; 600 - Three-dimensional optical neural network; 700 - Diffraction element. Detailed implementation manners

[0034] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this application will be more thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The drawings are merely schematic illustrations of the present application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repeated description will be omitted.

[0035] In addition, the described features, structures, or characteristics can be combined in any suitable manner in one or more example embodiments. In the following description, numerous specific details are provided to give a thorough understanding of the example embodiments of the present application. However, those skilled in the art will realize that the technical solutions of the present application can be practiced without one or more of the specific details, or other modules, components, etc. can be used. In other cases, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring the various aspects of the present application.

[0036] Optical neural networks use photons as information carriers and implement the transmission, modulation, and processing of optical signals through optical elements (such as modulators, optical waveguides, photodetectors). Optical neural networks combine optical elements and machine learning algorithms and aim to achieve more efficient computing through the physical characteristics of light (such as high speed, parallelism, and low energy consumption). Optical neural networks show potential in various computing tasks, especially in large-scale data processing tasks.

[0037] However, due to current hardware and technological limitations, insufficient integration of hardware such as photon chips results in a limited number of parallel channels, and it is difficult to balance the modulation efficiency and bandwidth of existing modulators (such as Mach-Zehnder modulators), leading to losses and errors during modulation. Technologically, existing deep learning algorithms are not optimized for the multi-dimensional characteristics of light, resulting in waste of hardware resources. All these lead to a relatively low input information density of the current optical neural network, with the optical neural network being able to process a small amount of data per unit time, thereby limiting the practical application scope of the optical neural network and making the optical neural network only suitable for simple optical computing tasks and difficult to meet the requirements of large-scale optical computing tasks.

[0038] In consideration of overcoming the above-mentioned defects of the related technologies, the present application provides an encoding module, an input port, and a three-dimensional input port for an optical neural network. Before information encoding, the present application converts the light beam into a plane wave, avoiding the influence of the original wavefront curvature of the light beam on the accuracy of information encoding and resulting in a decrease in the accuracy of the final optical computing task. Moreover, by using a tunable metasurface as the information encoder, the number of information that can be carried per unit area is increased, improving the input information density of the optical neural network, enabling the optical neural network to process more data streams, and thus being able to process higher-dimensional or more complex data structures, so that the optical neural network can handle large-scale optical computing tasks.

[0039] The present application provides an encoding module 100 for an optical neural network, as Figure 1 shown, Figure 1 FIG. shows a schematic diagram of the encoding module 100 provided by an embodiment of the present application. The encoding module 100 includes: a plane wave generator 101 and an information encoder 102. Among them, the plane wave generator 101 is arranged upstream of the information encoder 102 along the optical path propagation direction, Figure 1 and the arrow direction in Figure 1 is the optical path propagation direction. The plane wave generator is used to convert the received light beam (such as Figure 1 the arrow pointing to the plane wave generator 101) into a plane wave 200 and emit the plane wave 200 to the information encoder 102. After receiving the plane wave 200, the information encoder 102 encodes the plane wave 200. It should be noted that a plane wave is an electromagnetic wave whose wavefront is a plane during propagation.

[0040] Moreover, in the embodiments of the present application, the information encoder 102 is a reconfigurable metasurface. The reconfigurable metasurface includes at least two reconfigurable micro-nano units and can also be referred to as a tunable metasurface. The reconfigurable metasurface refers to a metasurface based on tunable materials. When the tunable materials are subjected to external excitation, the equivalent refractive index changes, thereby causing the phase at each position on the reconfigurable metasurface to change, realizing the dynamic regulation of physical parameters such as the phase and amplitude of light. The information encoder 102 can achieve phase regulation of N data points, so that an information vector of size 1×N can be encoded onto the optical field of the plane wave 200 and used as the input of the subsequent optical neural network 500.

[0041] For example, as Figure 2 shown, Figure 2 FIG. shows a working schematic diagram of the encoding module 100 provided by an embodiment of the present application. The plane wave generator 101 converts the received light beam into a plane wave 200. After the plane wave 200 propagates a certain distance, it reaches the information encoder 102. The reconfigurable metasurface corresponding to the information encoder 102 has adjusted the tunable materials at each position thereon according to the externally input information vector. Then, after receiving the plane wave 200, the information encoder 102 modulates the plane wave 200 (as shown by φ in Figure 2 , and the φ corresponds to the externally input information vector), loads the information corresponding to the information vector onto the optical field of the plane wave 200, and then inputs the modulated plane wave 200 into the calculation layer of the optical neural network 500.

[0042] In one embodiment, the calculation layer of the optical neural network 500 is implemented in a diffractive manner, that is, the calculation layer of the optical neural network 500 is one or more diffractive layers, and the optical neural network 500 is a diffractive neural network. However, it should be noted that this is only for the subsequent description of the present application and does not mean that the optical neural network 500 of the present application is limited to a diffractive neural network. It should be noted that the size of the effective area of the reconfigurable metasurface is determined by the size of the input information required by the calculation layer of the optical neural network 500. The effective area refers to the area on the reconfigurable metasurface where the light beam can be modulated.

[0043] In one embodiment, the reconfigurable micro-nano unit is composed of a nanostructure made of a tunable material and a corresponding substrate. When the tunable material is subjected to external excitation, the equivalent refractive index of the tunable material changes. As Figure 3 shown, Figure 3 FIG. shows a schematic diagram of the reconfigurable metasurface provided by an embodiment of the present application. Figure 3The dashed box therein represents a reconfigurable micro-nano unit. The dark squares in the dashed box represent nanostructures, and the substrate bears the tunable material. In this case, using the reconfigurable metasurface as the information encoder 102 can adjust the equivalent refractive index of the tunable material at each position on the reconfigurable metasurface in real time through external excitation. When processing different optical computing tasks, the coding method can be dynamically changed, improving the flexibility and adaptability of the optical neural network, and bringing better performance and a wider application range to the optical neural network.

[0044] As an example, the tunable material of the reconfigurable micro-nano unit can be a phase change material. When the phase change material is subjected to external excitation (such as heat, laser, applied voltage), it will change the internal lattice of the material, and can greatly change the dielectric constant of the material. Commonly used phase change materials include GST (germanium antimony telluride) and VO2 (vanadium dioxide).

[0045] Furthermore, in one embodiment, a reconfigurable micro-nano unit on the reconfigurable metasurface constitutes a data point. A data point refers to the smallest information unit in the information vector dataset that can be processed independently. Each data point performs phase modulation on the light beam received at the corresponding position according to the input information vector, and then loads the information corresponding to the information vector onto the optical field of the plane wave. That is, when there are M reconfigurable micro-nano units on the reconfigurable metasurface, the information encoder 102 can implement the phase modulation of M data points. In this case, due to the small size of the reconfigurable micro-nano units on the reconfigurable metasurface, the size of a single reconfigurable micro-nano unit can be controlled within 100 - 500 nm, and a single reconfigurable micro-nano unit can carry a piece of information as a data point. Therefore, more information can be carried in a limited space, enabling the information encoder 102 to maximize the input information density of the optical neural network, and thus maximizing the information processing ability of the optical neural network, and also enabling the optical neural network to meet the requirements of large-scale optical computing tasks.

[0046] In another embodiment, at least two reconfigurable micro-nano units on the reconfigurable metasurface constitute a data point, that is, multiple reconfigurable micro-nano units on the reconfigurable metasurface form a data point as a whole, and then perform phase modulation on the light beam received by this data point. In this case, while taking into account the dynamic regulation difficulty of the tunable metasurface, the information carrying capacity of the reconfigurable metasurface can be guaranteed, and the input information density of the optical neural network is also greatly improved.

[0047] In one embodiment, the external excitation can be one or more of optical excitation, electrical excitation, thermal excitation, and mechanical excitation.

[0048] Continuing with the above example, if the phase change material GST is used as the tunable material, the external excitations that can be used are thermal excitation or optical excitation. It should be noted that different external excitations have their own advantages. For example, when using optical excitation, there is no need for physical contact with the tunable material, avoiding mechanical interference, and the light spot can be focused to the nanoscale, which can meet the regulation of small-size tunable materials; when using electrical excitation, the voltage and current are easily digitally regulated, facilitating relevant switching, and the integration difficulty of electronic devices is low, and they can be directly integrated with the reconfigurable metasurface, thereby improving the integration degree of the optical neural network; when using thermal excitation, most phase change materials are sensitive to temperature, have strong adaptability, and have a low risk of causing irreversible damage to the tunable material, and are suitable for long-term stable operation; when using mechanical excitation, the phase change triggered by mechanical deformation is not affected by temperature, laser, or current, and is suitable for occasions with complex working environments.

[0049] In one embodiment, considering both the integration degree and computational efficiency of the optical neural network, this embodiment preferably uses optical excitation or electrical excitation to regulate the tunable material. In this case, the tunable material on the information encoder 102 provided in the present application can be regulated more quickly, so that the optical neural network can receive input information more quickly and perform subsequent computational tasks, improving the computational efficiency of the optical computing task.

[0050] In one embodiment, the plane wave generator 102 is composed of at least one metasurface and / or at least one refractive lens. That is to say, the plane wave generator 102 can be composed of at least one metasurface, or can be composed of at least one refractive lens, or can also be composed of a combination of at least one metasurface and at least one refractive lens.

[0051] As Figure 1 The plane wave generator 102 shown is the plane wave generator 102 composed of one metasurface. In this case, the phase on this metasurface is the collimation phase, so as to collimate the light beam received by this metasurface. An example of the plane wave generator 102 being a refractive lens is not shown in the drawings of the present application.

[0052] It should be noted that since the volume size of the encoding module 100 needs to be considered in practical applications, the plane wave generator 102 is preferably composed of one metasurface. A single metasurface can realize the conversion of the light beam into a plane wave. In this case, the volume of the encoding module can be minimized, thereby improving the integration degree of the optical chip and being more conducive to the large-scale layout of the optical chip.

[0053] The present application also provides an input port 300 for an optical neural network, as Figure 4 shown, Figure 4The figure shows a schematic diagram of the input port 300 provided by an embodiment of the present application. The input port 300 includes: an encoding module 100 as described in any of the above embodiments; an incident coupler 301; and an output coupler 302.

[0054] Specifically, the incident coupler 301, the output coupler 302, the plane wave generator 101 and the information encoder 102 of the encoding module 100 are arranged in sequence along the optical path propagation direction. The incident coupler 301 is connected to the output coupler 302 through an optical waveguide. The incident coupler 301 is used to receive the light beam emitted by the light source (as shown by the arrow pointing to the incident coupler 301 in Figure 4 ). The light beam propagates through the optical waveguide to the output coupler 302, and the output coupler 302 emits the light beam to the plane wave generator 102 of the encoding module 100.

[0055] It should be noted that the light beam emitted by the light source received by the incident coupler 301 is a converging light beam, and the converging point of the light beam is located on the incident coupler 301. The light beam output by the output coupler 302 is a diverging light beam, as shown in Figure 5 shown. Figure 5 The figure shows a schematic diagram of the input port 300 provided by an embodiment of the present application. The light beam emitted by the output coupler 302 is a diverging light beam. The plane wave generator 101 receives this diverging light beam and modulates it to generate a plane wave 200.

[0056] Furthermore, the incident coupler 301, the output coupler 302 and the optical waveguide connecting the two in the input port 300 are all made of materials with high transmittance in the working band. For example, when the working center wavelength is in the communication band of 1550 nm, the incident coupler 301, the output coupler 302 and the optical waveguide connecting the two are made of polysilicon, and the substrate carrying the three can be quartz glass. Note that this is only an example and does not mean that the working band of the input port 300 provided by the present application is limited to this. In fact, it can be one or more of the visible light band, the near-infrared band and the terahertz band, or other working bands, which are not limited in the present application.

[0057] The optical neural network 500 in the present application can be regarded as a two-dimensional optical neural network, and multiple optical neural networks 500 can form a three-dimensional optical neural network 600; the three-dimensional optical neural network can execute multiple tasks in parallel. It should be noted that the two-dimensional optical neural network mentioned here is mainly a two-dimensional abstract expression used to describe the connection method between the input port 100 and the computing layer of the optical neural network 500, and does not mean that the optical signal in the two-dimensional optical neural network propagates in a two-dimensional space; obviously, the optical signal in the two-dimensional optical neural network propagates in a three-dimensional space.

[0058] To this end, the present application also provides a three-dimensional input port 400 for a three-dimensional optical neural network, as Figure 6 shown, Figure 6 FIG. shows a schematic diagram of the three-dimensional input port 400 provided by an embodiment of the present application. The three-dimensional input port 400 includes at least two input ports 300 as described in the above embodiment.

[0059] Since the three-dimensional input port 400 is composed of multiple input ports 300, and each input port 300 also corresponds to an optical neural network 500 in the three-dimensional optical neural network 600. That is to say, during the actual execution of the optical computing task by the three-dimensional optical neural network 600, each input port 300 in the three-dimensional input port 400 can perform the same / different information encoding processing on the respective plane waves 200, and thus can execute multiple optical computing tasks in parallel.

[0060] As an example, as Figure 6 shown, Figure 6 each input port 300 in the three-dimensional input port 400 is denoted as the first port, the second port, and the third port from top to bottom, and each optical neural network 500 in the three-dimensional optical neural network 600 in Figure 6 is denoted as the first optical neural network, the second optical neural network, and the third optical neural network from top to bottom. The first port, the second port, and the third port are respectively in one-to-one correspondence with the first optical neural network, the second optical neural network, and the third optical neural network. The first port is used to execute the first task, the second port is used to execute the second task, and the third port is used to execute the third task. Then, during the actual optical computing process, the information encoder in the first port can perform information encoding on the plane wave of the first port, and load the information corresponding to the first task onto the plane optical field. The same applies to the second port and the third port. In this case, information encoding processing can be performed on different plane waves according to the actual optical computing task requirements, and thus the parallelism characteristic of light can be more fully utilized to achieve parallel processing of the three-dimensional optical neural network and improve the flexibility and computing efficiency of the three-dimensional optical neural network.

[0061] It should be noted that in this embodiment, each input port included in the three-dimensional input port has a corresponding light source in one-to-one correspondence, and each light source is used to emit a light beam to the incident coupler of the corresponding input port; and, similar to the two-dimensional optical neural network, the light beams emitted by each light source are convergent, and the convergence points are respectively on the corresponding incident coupler 301.

[0062] In one embodiment, the three-dimensional input port further includes a diffraction element 700, as Figure 7 shown, Figure 7FIG. 0 shows a schematic diagram of a three-dimensional input port 400 provided by an embodiment of the present application. The diffraction element 700 is configured to converge the light beams emitted by the light source to the incident couplers 301 of each input port in the three-dimensional input port 400. In this case, the entire three-dimensional optical neural network 600 can be driven by a single light source, avoiding the complex optical path alignment problems and synchronous / asynchronous control problems brought by using a light source array, reducing the complexity of the system, improving the stability and consistency of the system, and also reducing the manufacturing cost of the three-dimensional input port.

[0063] It should be noted that, in one embodiment, the light beam emitted by the light source to the diffraction element 700 is a collimated laser.

[0064] In one embodiment, the diffraction element 700 can be a metasurface or a diffraction grating. It should be noted that when using a diffraction grating as the diffraction element 700, due to the mature process of the diffraction grating, when the beam splitting requirement is relatively low, it is relatively convenient to design and manufacture a diffraction grating that meets the beam splitting requirement, reducing the manufacturing difficulty of the three-dimensional optical neural network; when using a metasurface as the diffraction element 700, due to the characteristics of the metasurface such as low manufacturing cost, nanoscale volume, and high design freedom, the cost of mass-producing the optical neural network can be reduced, and while controlling the cost, the volume of the optical neural network can also be reduced, and the periodic limitation of the grating can be broken, supporting complex beam splitting patterns (such as asymmetric beam splitting, custom energy distribution), thereby improving the flexibility of the arrangement of each optical neural network in the three-dimensional optical neural network, and also improving the integration degree of the three-dimensional optical neural network, and further reducing the volume of the three-dimensional optical neural network.

[0065] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the appended claims.

Claims

1. A coding module for an optical neural network, characterized in that: The encoding module includes: a plane wave generator; an information encoder; The plane wave generator is arranged upstream of the information encoder along the propagation direction of the optical path; The plane wave generator is used to convert the received light beam into a plane wave, and the information encoder is used to encode information on the plane wave; Wherein, the information encoder is a reconfigurable metasurface, and the reconfigurable metasurface includes at least two reconfigurable micro-nano units.

2. The encoding module according to claim 1, characterized in that: A reconfigurable micro-nano unit on the reconfigurable metasurface constitutes a data point.

3. The encoding module according to claim 1, characterized in that: At least two reconfigurable micro-nano units on the reconfigurable metasurface constitute a data point.

4. The encoding module according to claim 1, characterized in that: The plane wave generator is composed of at least one metasurface and / or at least one refractive lens.

5. The encoding module according to claim 1, characterized in that: The equivalent refractive index of the reconfigurable micro-nano unit will change when subjected to external stimulation.

6. The encoding module according to claim 5, characterized in that: The external excitation is one or more of optical excitation, electrical excitation, thermal excitation, and mechanical excitation.

7. An input port for an optical neural network, characterized in that The input port comprises: the encoding module according to any one of claims 1 to 6; an input coupler; an output coupler; Wherein, the incident coupler and the output coupler are connected via an optical waveguide; The incident coupler is used to receive the light beam emitted by the light source, and the output coupler is used to emit the light beam to the encoding module.

8. A three-dimensional input port for a three-dimensional optical neural network, characterized in that: The three-dimensional input port comprises: at least two input ports as claimed in claim 7.

9. The three-dimensional input port according to claim 8, characterized in that: The three-dimensional input port further includes: a diffraction element; The diffraction element is used to converge the light beams emitted by the light source onto the input couplers of each of the input ports.

10. The three-dimensional input port according to claim 9, characterized in that: The diffraction element is a super surface or a diffraction grating.

Images