Intelligent optical computing general propagation model, architecture and system

Abstract

The invention relates to the technical field of optical computing, in particular to an intelligent optical computing general propagation model, architecture and system. The model comprises a model training module which is used for obtaining a target light calculation task and training an initial light calculation general propagation model according to the target light calculation task to obtain a target light calculation general propagation model, the initial light calculation general propagation model is constructed by a four-dimensional wave optical light field function, and the target light calculation general propagation model is constructed by a four-dimensional wave optical light field function; the four-dimensional fluctuation optical light field function is used for describing information of a plurality of light fields in a spatial dimension, a spectral dimension and a polarization dimension; and the optical calculation module is used for acquiring an optical input signal, modulating the optical input signal based on the target optical calculation general propagation model to obtain a modulated optical signal, and determining an optical calculation result according to the modulated optical signal. By adopting the scheme, the flexibility and the practicability of optical calculation can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of optical computing technology, and in particular to a general propagation model, architecture and system for intelligent optical computing. Background Art

[0002] With the rapid development of artificial intelligence and scientific computing, the complexity and scale of computing needs are also increasing. However, the existing electronic computing technology is limited by Moore's Law, and its performance is gradually approaching saturation, making it difficult to effectively cope with the increasingly stringent requirements of large-scale complex algorithms on computing power and power consumption. Light has natural advantages such as high throughput and low latency in the propagation process. Optical computing technology that uses photons instead of electrons as computing carriers is seen as the key to breaking the existing computing bottleneck. Summary of the invention

[0003] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0004] To this end, a first objective of the present disclosure is to propose a universal propagation model for intelligent optical computing to improve the flexibility and practicality of optical computing.

[0005] The second objective of the present disclosure is to propose a universal propagation architecture for intelligent optical computing.

[0006] The third objective of the present disclosure is to propose a universal propagation system for intelligent optical computing.

[0007] To achieve the above objectives, the first embodiment of the present disclosure proposes a general propagation model for intelligent optical computing, including: A model training module, used to obtain a target light computing task, and train an initial light computing universal propagation model according to the target light computing task to obtain a target light computing universal propagation model, wherein the initial light computing universal propagation model is constructed by a four-dimensional wave optics light field function, and the four-dimensional wave optics light field function is used to describe the information of the complex light field in the spatial dimension, the spectral dimension, and the polarization dimension; The optical computing module is used to obtain an optical input signal, and modulate the optical input signal based on the target optical computing universal propagation model to obtain a modulated optical signal, so as to determine an optical computing result according to the modulated optical signal.

[0008] Optionally, the initial light computing universal propagation model includes a first initial light computing universal propagation sub-model of an interference dimension, and before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: A first initial light calculation universal propagation sub-model of interference dimension is constructed, wherein the first initial light calculation universal propagation sub-model is obtained by adding a first four-dimensional wave optics light field function and a second four-dimensional wave optics light field function.

[0009] Optionally, the initial light computation universal propagation model includes a second initial light computation universal propagation sub-model in a diffraction dimension, and before training the initial light computation universal propagation model according to the target light computation task, the model training module is further used to: A second initial light calculation universal propagation sub-model of the diffraction dimension is constructed, wherein the second initial light calculation universal propagation sub-model is obtained by convolving a third four-dimensional wave optical light field function with a diffraction transfer function corresponding to the third four-dimensional wave optical light field function.

[0010] Optionally, the initial light computing universal propagation model includes a third initial light computing universal propagation sub-model of a polarization dimension, and before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: A third initial light calculation universal propagation sub-model of polarization dimension is constructed, wherein the third initial light calculation universal propagation sub-model is obtained by performing Jones matrix multiplication operation on the Jones vector of the fourth four-dimensional wave optical light field function in each wavelength channel.

[0011] Optionally, the initial light computing universal propagation model includes a fourth initial light computing universal propagation sub-model of a spectral dimension, and before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: A fourth initial light calculation universal propagation sub-model of spectral dimension is constructed, wherein the fourth initial light calculation universal propagation sub-model is obtained by filtering and integrating information of spectral dimension and polarization dimension of the fifth four-dimensional wave optical light field function.

[0012] Optionally, when the optical computing module is used to modulate the optical input signal based on the target optical computing universal propagation model, it is specifically used to: Adjusting the structural parameters of the initial metasurface structure based on the target light calculation universal propagation model to obtain a target metasurface structure; The optical input signal is input into the target metasurface structure to control the target metasurface structure to modulate the optical input signal based on the target light calculation universal propagation model.

[0013] Optionally, each pixel on the initial metasurface structure includes a plurality of rectangular nanocolumns, and the optical computing module is used to adjust the structural parameters of the initial metasurface structure based on the target optical computing universal propagation model, specifically for: Calculating a universal propagation model of the target light to determine adjustment parameters corresponding to each of the rectangular nanocolumns, wherein the adjustment parameters include a rotation angle, an amplitude modulation coefficient, a phase modulation coefficient, and a center offset; Based on the adjustment parameters, the rectangular nanorods are adjusted.

[0014] Optionally, when the model training module is used to train the initial light computing universal propagation model according to the target light computing task, it is specifically used to: Constructing a loss function corresponding to the target light computing task; Based on the loss function, a gradient descent method is used to train model parameters in the initial light computation universal propagation model.

[0015] To achieve the above-mentioned purpose, the second aspect of the present disclosure proposes a general propagation architecture for intelligent optical computing, including: a data input module, a general propagation model for intelligent optical computing as shown in any one of the first aspects, and an output acquisition module; wherein: The data input module is used to obtain input data and encode the input data into a spatial dimension, a spectral dimension, and a polarization dimension to obtain an optical input signal; The intelligent optical computing universal propagation model is used to obtain an optical input signal, and modulate the optical input signal based on the target optical computing universal propagation model corresponding to the target optical computing task to obtain a modulated optical signal; The output acquisition module is used to collect the modulated optical signal to obtain and output an optical calculation result signal.

[0016] To achieve the above-mentioned purpose, the third aspect of the present disclosure proposes an intelligent optical computing universal propagation system, including: the intelligent optical computing universal propagation architecture shown in the second aspect above.

[0017] In summary, the general propagation model, architecture and system of intelligent optical computing provided by the present invention, by constructing an initial general propagation model of optical computing based on a four-dimensional wave optical light field function, can provide a general light field propagation theory and modeling method for optical computing and optical neural networks, thereby improving the flexibility and practicality of optical computing, improving the design efficiency and applicability of optical neural networks, and realizing wide applications in multiple fields.

[0018] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A schematic diagram of the structure of a general propagation model for intelligent optical computing provided by an embodiment of the present disclosure; Figure 2 A schematic diagram of the structure of a general propagation architecture for intelligent optical computing provided by an embodiment of the present disclosure; Figure 3 A schematic diagram of the structure of a universal propagation architecture for intelligent optical computing provided by another embodiment of the present disclosure. DETAILED DESCRIPTION

[0020] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0021] With the rapid development of artificial intelligence, how to improve computing speed and efficiency has become a key issue. Optical computing uses photons rather than electrons to process and calculate information, and has the advantages of high speed, low energy consumption and parallel processing. Optical computing technology relies on the physical properties of light such as wavelength, polarization, diffraction, and interference. For example, light interference forms interference patterns through superimposed light waves, which can be used to process information. Mach-Zehnder interferometers are often used for matrix multiplication in optical computing. The diffraction properties of light can realize parallel processing and calculation of data by controlling the propagation path of light waves in different media. Diffraction optical neural networks can realize tasks such as image classification at the speed of light. The polarization properties of light can be used for multiplexing of multiple tasks. However, the optical propagation principles and corresponding mathematical models involved in current optical computing technology are relatively scattered, independent of each other, lack close connections, and lack a universal propagation model for intelligent optical computing. This makes it difficult to effectively combine different optical computing technologies, support the design of highly complex optical computing systems, and adapt to diverse application requirements. The lack of unified standards and universal propagation models makes it difficult for different systems to be compatible, which restricts the popularization and development of optical computing technology in a wider range of computing applications.

[0022] The present disclosure is described in detail below with reference to specific embodiments.

[0023] Figure 1 This is a schematic diagram of the structure of a general propagation model for intelligent optical computing provided by an embodiment of the present disclosure. Figure 1 As shown, the intelligent optical computing universal propagation model includes: A model training module is used to obtain a target light computing task, and train an initial light computing universal propagation model according to the target light computing task to obtain a target light computing universal propagation model; The optical computing module is used to obtain an optical input signal and modulate the optical input signal based on a target optical computing universal propagation model to obtain a modulated optical signal, so as to determine an optical computing result according to the modulated optical signal.

[0024] According to some embodiments, the target optical computing task refers to an optical computing task that needs to be performed on the optical input signal. The target optical computing task does not specifically refer to a fixed task.

[0025] In some embodiments, the general propagation model for initial light calculation is constructed by a four-dimensional wave optics light field function. The four-dimensional wave optics light field function refers to a function constructed using wave optics theory. The four-dimensional wave optics light field function is used to describe the information of the complex light field in the spatial dimension, spectral dimension, and polarization dimension. The four-dimensional wave optics light field function reflects the correlation between the input data and the characteristics of the light source.

[0026] In some embodiments, the mathematical expression of the four-dimensional wave optics light field function can be ,in, , represents two spatial dimensions, represents the spectral dimension, Indicates that it contains , Polarization dimensions in two directions.

[0027] In some embodiments, the four-dimensional wave optics light field function can be used to construct a universal propagation sub-model for light computing in dimensions such as interference, diffraction, polarization, and spectrum.

[0028] It should be noted that the universal propagation model of intelligent optical computing provided by the embodiments of the present disclosure, by constructing an initial universal propagation model of optical computing based on a four-dimensional wave optical light field function, can provide a universal light field propagation theory and modeling method for optical computing and optical neural networks, thereby improving the flexibility and practicality of optical computing, improving the design efficiency and applicability of optical neural networks, and realizing wide applications in multiple fields.

[0029] Optionally, the initial light computing universal propagation model includes a first initial light computing universal propagation sub-model of the interference dimension. Before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: Construct the first initial light calculation universal propagation sub-model in interference dimension.

[0030] It should be noted that the interference of the two light paths can be expressed as the direct addition of two four-dimensional light fields corresponding to each wavelength and polarization channel, that is, .

[0031] That is, the first initial light calculation universal propagation submodel can be composed of the first four-dimensional wave optics light field function and the second four-dimensional wave optics light field function Add together.

[0032] According to some embodiments, modulation of an optical signal in an interference dimension may be achieved by using a spatial light modulator and calculating a general propagation sub-model based on a first initial light.

[0033] Optionally, the initial light computing universal propagation model includes a second initial light computing universal propagation sub-model in a diffraction dimension. Before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: A second initial light calculation universal propagation sub-model of the diffraction dimension is constructed, wherein the second initial light calculation universal propagation sub-model is obtained by convolving the third four-dimensional wave optical light field function with the diffraction transfer function corresponding to the third four-dimensional wave optical light field function.

[0034] For example, the diffraction of light can be represented as the third four-dimensional light field at z=0 , each wavelength, polarization channel is independently related to the distance d The corresponding diffraction transfer function Perform convolution, that is:

[0035] in, Indicates j Wavelength, k Rayleigh-Sommerfeld impulse response function under polarization channels; i Represents the imaginary number symbol; r Indicates distance; It is also possible to use the properties of Fourier transform to convert it into frequency domain calculation, namely:

[0036] in,

[0037] According to some embodiments, the modulation of the optical signal in the interference dimension may be achieved by using a diffractive optical element and calculating a general propagation sub-model based on the second initial light.

[0038] Optionally, the initial light computing universal propagation model includes a third initial light computing universal propagation sub-model of a polarization dimension, and before the initial light computing universal propagation model is trained according to the target light computing task, the model training module is further used to: A third initial light calculation universal propagation sub-model of polarization dimension is constructed, wherein the third initial light calculation universal propagation sub-model is obtained by performing Jones matrix multiplication operation on the Jones vector of the fourth four-dimensional wave optical light field function in each wavelength channel.

[0039] According to some embodiments, the Jones matrix multiplication operation may be performed according to the following formula:

[0040] in, Represents the polarization, intensity, and phase modulation effects of each pixel in the spatial dimension of the fourth 4D wave optical light field function.

[0041] According to some embodiments, polarization, intensity, and phase modulation of an optical input signal may be achieved through a metasurface micro-nanostructure.

[0042] That is to say, the optical computing module can adjust the structural parameters of the initial metasurface structure based on the target optical computing general propagation model to obtain the target metasurface structure; and input the optical input signal into the target metasurface structure to control the target metasurface structure to modulate the optical input signal based on the target optical computing general propagation model.

[0043] In some embodiments, each pixel on the initial metasurface structure includes multiple rectangular nanocolumns, and the optical computing module can determine the adjustment parameters corresponding to each rectangular nanocolumn based on the target light computing general propagation model; and adjust the rectangular nanocolumns based on the adjustment parameters.

[0044] In some embodiments, the adjustment parameters include rotation angle, amplitude modulation coefficient, phase modulation coefficient, and center offset. The adjustment parameters can form a Jones matrix, as shown in the following formula:

[0045] in, Indicated by the rotation angle The rotation matrix is ​​formed, , and , represent the amplitude modulation coefficient and phase modulation coefficient in the length and width directions of the rectangular nanorod, respectively. , represents the center offset of the rectangular nanorod.

[0046] The Jones matrix of the entire pixel containing multiple rectangular nanopillars can be expressed as The adjustment parameters of multiple rectangular nanorods are independent of each other. Therefore, four rectangular nanorods can realize a Jones matrix with full degrees of freedom, achieving arbitrary polarization, amplitude, and phase modulation. Similarly, rectangular nanorods respond differently to different wavelengths, and complete spectral modulation can be achieved through special design.

[0047] Optionally, the initial light computing universal propagation model includes a fourth initial light computing universal propagation sub-model of the spectral dimension, and before the initial light computing universal propagation model is trained according to the target light computing task, the model training module is further used to: Construct the fourth initial light computation universal propagation sub-model in spectral dimension.

[0048] According to some embodiments, the fourth initial light calculation universal propagation submodel can be used to indicate the intensity of the entire light field. That is, the fourth initial light calculation can be obtained by filtering and integrating the information of the spectral dimension and the polarization dimension of the fifth four-dimensional wave optical light field function, as shown in the following formula:

[0049] in, Indicates the detection sensitivity of the sensor for different wavelengths and polarizations.

[0050] According to some embodiments, modulation of an optical signal in a spectral dimension may be achieved by using a digital micromirror device based on a fourth initial optical computation universal propagation submodel.

[0051] Optionally, when the model training module is used to train the initial light computing universal propagation model according to the target light computing task, it is specifically used to: Construct the loss function corresponding to the target optical computing task; Based on the loss function, the gradient descent method is used to train the model parameters in the initial light computation general propagation model.

[0052] It should be noted that, based on the loss function, other model training methods other than the gradient descent method may be used to train the model parameters in the initial light calculation general propagation model.

[0053] In order to implement the above embodiments, the present disclosure also proposes a general propagation architecture for intelligent optical computing.

[0054] like Figure 2 As shown, the intelligent optical computing universal propagation architecture includes: a data input module, the intelligent optical computing universal propagation model provided in the above embodiments, and an output acquisition module; wherein, A data input module is used to obtain input data and encode the input data into a spatial dimension, a spectral dimension, and a polarization dimension to obtain an optical input signal; The intelligent optical computing universal propagation model is used to obtain an optical input signal, and modulate the optical input signal based on the target optical computing universal propagation model corresponding to the target optical computing task to obtain a modulated optical signal; The output acquisition module is used to collect the modulated optical signal, obtain and output the optical calculation result signal.

[0055] According to some embodiments, the data input module can encode the input data into the spatial dimension, spectral dimension, and polarization dimension through devices such as a spatial light modulator, a mask, a digital micromirror device, and a multi-wavelength illumination light source. In particular, the encoding of the intensity dimension and the phase dimension can also be realized on the basis of the spatial dimension, the spectral dimension, and the polarization dimension.

[0056] In some embodiments, the output acquisition module can detect the modulated light signal through sensors such as photodetectors and cameras.

[0057] Take a scenario as an example. Figure 3 This is a schematic diagram of a general propagation architecture for intelligent optical computing provided by another embodiment of the present disclosure. Figure 3 As shown in the figure, it uses multi-layer metasurfaces to realize multi-dimensional light field modulation of the general propagation model of intelligent optical computing. Specifically, by combining the interference, diffraction, polarization, and spectral calculation methods involved in the general propagation model of intelligent optical computing, it can build a complete numerical propagation model from the input plane to the output plane, and then combine it with artificial intelligence algorithms to design and optimize the initial metasurface structure to build a general propagation architecture for intelligent optical computing for specific target optical computing tasks.

[0058] In summary, the architecture provided in this embodiment establishes a numerical model from input to output by using a general propagation model for optical calculation based on spectrum, polarization, space and other dimensions, and can perform multi-dimensional modulation of the light field. After training, it can efficiently execute corresponding artificial intelligence tasks.

[0059] In order to implement the above embodiments, the present disclosure also proposes an intelligent optical computing universal propagation system, including: the intelligent optical computing universal propagation architecture provided by the above embodiments.

[0060] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this disclosure shall comply with the relevant laws and regulations and shall not violate public order and good morals.

[0061] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0062] The present disclosure anticipates providing implementation schemes for users to selectively block the use or access of personal information data. That is, the present disclosure anticipates providing hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, risks can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0063] The acquisition, transmission, storage, use, and processing of data in the technical solution disclosed in this disclosure are in compliance with the relevant provisions of national laws and regulations.

[0064] It should be noted that in the embodiments of the present disclosure, certain software, components, models and other existing solutions in the industry may be mentioned, which should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0065] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they contradict each other.

[0066] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0067] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.

[0068] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing in a suitable manner if necessary, and then stored in a computer memory.

[0069] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0070] A person skilled in the art may understand that all or part of the steps in the above-mentioned embodiment method may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0071] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0072] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.

Claims

1. A general propagation model for intelligent optical computing, characterized in that: include: A model training module, used to obtain a target light computing task, and train an initial light computing universal propagation model according to the target light computing task to obtain a target light computing universal propagation model, wherein the initial light computing universal propagation model is constructed by a four-dimensional wave optics light field function, and the four-dimensional wave optics light field function is used to describe the information of the complex light field in the spatial dimension, the spectral dimension, and the polarization dimension; The optical computing module is used to obtain an optical input signal, and modulate the optical input signal based on the target optical computing universal propagation model to obtain a modulated optical signal, so as to determine an optical computing result according to the modulated optical signal.

2. The model according to claim 1, characterized in that The initial light computing universal propagation model includes a first initial light computing universal propagation sub-model of interference dimension. Before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: A first initial light calculation universal propagation sub-model of interference dimension is constructed, wherein the first initial light calculation universal propagation sub-model is obtained by adding a first four-dimensional wave optics light field function and a second four-dimensional wave optics light field function.

3. The model according to claim 1, characterized in that The initial light calculation universal propagation model includes a second initial light calculation universal propagation sub-model in a diffraction dimension. Before training the initial light calculation universal propagation model according to the target light calculation task, the model training module is further used to: A second initial light calculation universal propagation sub-model of the diffraction dimension is constructed, wherein the second initial light calculation universal propagation sub-model is obtained by convolving a third four-dimensional wave optical light field function with a diffraction transfer function corresponding to the third four-dimensional wave optical light field function.

4. The model according to claim 1, characterized in that The initial light computing universal propagation model includes a third initial light computing universal propagation sub-model of a polarization dimension. Before training the initial light computing universal propagation model according to the target light computing task, the model training module is further used to: A third initial light calculation universal propagation sub-model of polarization dimension is constructed, wherein the third initial light calculation universal propagation sub-model is obtained by performing Jones matrix multiplication operation on the Jones vector of the fourth four-dimensional wave optical light field function in each wavelength channel.

5. The model according to claim 1, characterized in that The initial light computing universal propagation model includes a fourth initial light computing universal propagation sub-model of the spectral dimension. Before the initial light computing universal propagation model is trained according to the target light computing task, the model training module is further used to: A fourth initial light calculation universal propagation sub-model of spectral dimension is constructed, wherein the fourth initial light calculation universal propagation sub-model is obtained by filtering and integrating information of spectral dimension and polarization dimension of the fifth four-dimensional wave optical light field function.

6. The model according to claim 1, characterized in that When the optical computing module is used to modulate the optical input signal based on the target optical computing universal propagation model, it is specifically used to: Adjusting the structural parameters of the initial metasurface structure based on the target light calculation universal propagation model to obtain a target metasurface structure; The optical input signal is input into the target metasurface structure to control the target metasurface structure to modulate the optical input signal based on the target light calculation universal propagation model.

7. The model according to claim 6, characterized in that Each pixel on the initial supersurface structure includes a plurality of rectangular nanocolumns, and the optical computing module is used to adjust the structural parameters of the initial supersurface structure based on the target optical computing universal propagation model, specifically for: Based on the target light calculation universal propagation model, determining the adjustment parameters corresponding to each of the rectangular nanocolumns, wherein the adjustment parameters include a rotation angle, an amplitude modulation coefficient, a phase modulation coefficient, and a center offset; Based on the adjustment parameters, the rectangular nanorods are adjusted.

8. The model according to claim 1, characterized in that When the model training module is used to train the initial light computing universal propagation model according to the target light computing task, it is specifically used to: Constructing a loss function corresponding to the target light computing task; Based on the loss function, a gradient descent method is used to train model parameters in the initial light computation universal propagation model.

9. A general propagation architecture for intelligent optical computing, characterized in that: include: A data input module, a general propagation model for intelligent optical computing according to any one of claims 1 to 8, and an output acquisition module; wherein: The data input module is used to obtain input data and encode the input data into a spatial dimension, a spectral dimension, and a polarization dimension to obtain an optical input signal; The intelligent optical computing universal propagation model is used to obtain an optical input signal, and modulate the optical input signal based on the target optical computing universal propagation model corresponding to the target optical computing task to obtain a modulated optical signal; The output acquisition module is used to collect the modulated optical signal to obtain and output an optical calculation result signal.

10. An intelligent optical computing universal propagation system, characterized in that: include: At least one intelligent optical computing universal propagation architecture as described in claim 9.

Images