Photon processor, data identification method, device, equipment, medium and product

Abstract

The invention discloses a photon processor, a data identification method, a data identification device, data identification equipment, a medium and a product, and relates to the field of optical calculation. Wherein the operation process simulation optical path comprises a plurality of tunable photonic devices which are deployed according to a one-time iterative operation process of a matrix to be calculated, and two optical path output channels of the loop iteration control optical path are respectively communicated to the input end of the operation process simulation optical path and the result reading optical path. An optical path switching control device transmits optical compensation signals output by previous k-1 iterative operations to an operation process simulation optical path for next iterative operations, optical compensation signals of the (k-1) th and kth iterative operations are transmitted to a result reading optical path, and the result reading optical path transmits the (k-1) th iterative operation result to the operation process simulation optical path. And obtaining feature data according to results of the two operations. The problem that photoelectric / electro-optical conversion needs to be repeated for multiple times in the prior art can be solved, the data processing efficiency is effectively improved, and resources are saved.

Description

Technical Field

[0001] The present invention relates to the field of optical computing, and in particular to a photon processor and a data recognition method, device, equipment, medium, and product. Background Art

[0002] With the explosive growth of data scale, the computational complexity of traditional matrix characteristic problems has increased dramatically. In order to improve the efficiency of matrix data calculation, photonic computing is currently used instead of traditional electronic computing. In the process of using photonic processors to process multiple iterative matrix operations, the relevant technology requires the intermediate results to be output from the optical domain to the electronic storage unit in each round of iterative operations. The repeated photoelectric conversion / electro-optical conversion leads to low overall data processing efficiency. Summary of the Invention

[0003] The present invention provides a photon processor and a target data identification method, device, electronic device, non-volatile storage medium, and computer program product, which can realize the calculation of matrix characteristic problems through a photon processor with a relatively small number of photoelectric / electro-optical conversions, effectively improve data processing efficiency, and save resources.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions: The present invention also provides a photonic processor comprising a computational process simulation optical path, a loop iteration control optical path, and a result reading optical path. The computational process simulation optical path comprises a plurality of tunable photonic devices deployed according to a single iterative computational process of a matrix to be computed, with the output end of the device connected to the input end of the loop iteration control optical path. The matrix to be computed undergoes k iterative computations to obtain characteristic data.

[0005] The first optical path output channel of the cyclic iteration control optical path is connected to the input end of the calculation process simulation optical path, and the second optical path output channel is connected to the result reading optical path. Its optical path switching control device transmits the optical compensation signals corresponding to the first k-1 iterative operations to the calculation process simulation optical path for the next iterative operation through the first optical path output channel, and transmits the optical compensation signals corresponding to the k-1th and kth iterative operations to the result reading optical path through the second optical path output channel.

[0006] The result reading optical path is connected to the input end of the operation process simulation optical path, and the first data signal corresponding to the k-1th iterative operation is transmitted to the operation process simulation optical path for the kth iterative operation; the characteristic data is obtained according to the first data signal and the second data signal corresponding to the kth iterative operation.

[0007] The present invention also provides a target data identification method, comprising: When a data identification task of identifying target data from task data is received, task data related to the data identification task is acquired.

[0008] The task data is expressed in matrix form to obtain the matrix to be calculated, and the characteristic data of the matrix to be calculated is the target data.

[0009] When it is determined that the characteristic data calculation process of the matrix to be calculated includes multiple iterative matrix-vector multiplication operations or multiple iterative matrix multiplication operations, the characteristic data of the matrix to be calculated is obtained by calling any of the above-mentioned photon processors to perform data processing on the matrix to be calculated as the processing result of the data recognition task.

[0010] The present invention also provides a target data identification device, comprising: The task data acquisition module is used to acquire task data related to the data identification task when receiving a data identification task of identifying target data from task data.

[0011] The data processing module is used to express the task data in matrix form to obtain the matrix to be calculated, and the characteristic data of the matrix to be calculated is the target data.

[0012] The task execution module is used to, when it is determined that the characteristic data calculation process of the matrix to be calculated includes multiple iterative matrix-vector multiplication operations or multiple iterative matrix multiplication operations, perform data processing on the matrix to be calculated by calling any of the above-mentioned photon processors to obtain the characteristic data of the matrix to be calculated as the processing result of the data recognition task.

[0013] The present invention also provides an electronic device comprising a memory and a processor, wherein the processor is configured to implement the steps of any of the above-mentioned target data identification methods when executing a computer program stored in the memory.

[0014] The present invention also provides a non-volatile storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of any of the above-mentioned target data identification methods are implemented.

[0015] Finally, the present invention also provides a computer program product, comprising a computer program / instruction, which implements the steps of any of the above target data identification methods when executed by a processor.

[0016] The advantage of the technical solution provided by the present invention is that, for a matrix whose characteristic data is calculated through multiple cyclic iterative processes, based on the mapping relationship between a tunable photonic device and the matrix elements, a group of tunable photonic devices can be used to simulate the single iterative operation process of the matrix, and the optical signal output of each iterative operation is compensated by the cyclic iterative control optical path, and the optical signal after the previous k-1 compensation is controlled to enter the operation process simulation optical path according to the current number of iterations to participate in the next iterative operation, and the optical compensation signals of the k-1th and kth times are output from the optical domain to the result reading module for optical-to-electrical conversion, written to the electronic storage unit, and then the output of the k-1th time is converted to the optical signal. The output is reconverted into an optical signal input loop iteration to control the optical path to participate in the last iteration. The whole process only requires 2 photoelectric / electro-optical conversions, which reduces k-2 photoelectric / electro-optical conversions compared with related technologies. It not only effectively avoids the increase in system delay caused by photoelectric signal conversion and improves the matrix data processing efficiency, but also avoids the energy consumption overhead caused by k-2 photoelectric / electro-optical conversions, saves resources, effectively reduces the power consumption of optical computing, effectively improves the computing power of the photonic processor, and avoids the problem of limited data transmission rate due to the sampling rate and bandwidth limitations of photoelectric / electro-optical conversion that cannot fully match the high-speed characteristics of optical computing, thereby further improving the matrix data processing efficiency.

[0017] In addition, the present invention also provides corresponding methods, devices, electronic devices, non-volatile storage media and computer program products for the photonic processor, further making the photonic processor more practical, and the target data recognition method, device, electronic device, non-volatile storage medium and computer program product have corresponding advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the present invention or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 A schematic diagram of the hardware composition framework applicable to the photonic processor provided by the present invention; Figure 2 A schematic diagram of the structural framework of a photon processor provided by the present invention in an exemplary embodiment; Figure 3 A schematic structural diagram of a photon processor provided by the present invention in another exemplary embodiment; Figure 4 A schematic structural diagram of a first type of photonic device provided by the present invention in an exemplary embodiment; Figure 5The matrix provided by the present invention represents a schematic structural diagram of an optical device array in an exemplary embodiment; Figure 6 A schematic structural diagram of a second type of photonic device provided by the present invention in an exemplary embodiment; Figure 7 A schematic structural diagram of an optical path selector provided by the present invention in an exemplary embodiment; Figure 8 A schematic diagram of the structure of a simulation optical path for a computation process provided by the present invention in an exemplary embodiment; Figure 9 A flow chart of a target data identification method provided by the present invention; Figure 10 A structural framework diagram of an exemplary embodiment of a target data identification device provided by the present invention; Figure 11 This is a structural diagram of an exemplary embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. The terms "first," "second," "third," "fourth," etc. in the specification and the accompanying drawings are used to distinguish different objects rather than to describe a specific order. Furthermore, the terms "including" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. The term "exemplary" means "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as superior or better than other embodiments.

[0021] With the rapid development of artificial intelligence technology and big data, the scale of data has exploded, such as high-dimensional images, large-scale graph data, and deep learning parameter matrices. Traditional matrix feature problem calculation methods, such as the QR (orthogonal matrix integral solution) algorithm, have a sharp increase in computational complexity for large-scale matrix data. For example, for a matrix with a dimension of The complexity of the classical algorithm is usually as high as , which makes it difficult to meet user needs in terms of efficiency when processing millions or even larger amounts of data.

[0022] To overcome the physical limitations of traditional electronic processors due to Moore's Law and the limitations of the von Neumann architecture, and to address the computing power and functionality bottlenecks of current classical computers, optical computing applications, a computing paradigm based on photon signals as the fundamental carrier for information transmission and processing, have emerged. Optical computing leverages the ultra-high-speed transmission and interference properties of light waves to perform rapid calculations, offering advantages such as large bandwidth, ultra-high speed, low power consumption, low latency, and high parallelism. Photonic processors based on optical computing, with their inherent high parallelism, low latency, and low energy consumption, are being applied in fields with high computational complexity. For example, photonic processors can be used to perform matrix-vector multiplication operations in image classification and speech recognition tasks.

[0023] In practical applications, such as data science, physical modeling, and engineering optimization, when solving high-dimensional, dynamic, and large-scale problems, the eigenvalues ​​of matrices require multiple iterations of matrix operations to meet subsequent processing requirements. This approach offers the advantage of balancing accuracy and efficiency, sacrificing some mathematical rigor in exchange for computational feasibility, resource conservation, and real-time responsiveness in real-world scenarios. Currently, in implementing such iterative algorithms, photonic processors in related technologies must output intermediate results from the optical domain to electronic storage units at each iteration, which are then processed and reconverted back into optical signal inputs. Calculating a single eigenvalue of a matrix requires repeated photoelectric / electro-optical conversions. This high-speed, high-precision photoelectric conversion process typically requires nanoseconds, far exceeding the picosecond propagation speed of optical signals within the photonic processor. This makes photoelectric conversion a bottleneck in the data processing efficiency of the entire photonic processor. Furthermore, frequent photoelectric signal conversions not only increase the data processing latency of the photonic processor but also impose additional energy consumption, partially offsetting the inherent low power consumption advantage of optical computing. At the same time, some sampling rates and bandwidth limitations of optoelectronic / electro-optical conversions may not fully match the high-speed characteristics of optical computing, resulting in limited data transmission rates and affecting the overall computing efficiency of the photonic processor.

[0024] In view of this, in order to solve the above problems, the present invention is based on the mapping relationship between a tunable photonic device and a matrix element, and uses a group of tunable photonic devices to simulate a single iterative process of the matrix. The optical signal output of each iterative operation is compensated by the cyclic iterative control optical path, and the opening of the two optical path output channels is controlled according to the current number of iterations. The optical signal after the first k-1 compensations is transmitted into the operation process simulation optical path to participate in the next iterative operation, and the optical compensation signal corresponding to the k-1th iterative operation process is transmitted to the result reading circuit. The k-1th optical compensation signal of the result reading circuit is reconverted into an optical signal input cyclic iterative control optical path to participate in the last iteration, and the optical compensation signal corresponding to the kth iterative operation process is transmitted to the result reading circuit for result reading. The entire process only requires 2 result reading circuits to perform 2 photoelectric / electro-optical conversions, thereby realizing the calculation of the matrix characteristic problem with a photonic processor with fewer photoelectric / electro-optical conversions, so as to achieve the purpose of saving resources and improving computing power.

[0025] In conjunction with the specific application environment architecture or specific hardware architecture on which the execution of the matrix feature data calculation process depends, the specific application environment architecture or specific hardware architecture is described here. Figure 1 Some possible application scenarios involved in the technical solution of the present invention are introduced by way of example, which may include the following: An artificial intelligence platform is deployed on the photonic computer 1. The artificial intelligence platform performs artificial intelligence tasks that require calculating the eigenvalues ​​of matrix data, which may include data dimensionality reduction tasks, pattern recognition tasks, network analysis tasks, and model optimization tasks, through the photonic processor 10 of the photonic computer 1. A photonic computer is a new type of computer that performs digital calculations, logical operations, information storage, and processing by optical signals. It is composed of at least optical elements and devices such as lasers, optical reflectors, lenses, and filters. Information processing is performed by entering an array of reflectors and lenses through a laser beam, replacing electrons with photons, and realizing optical calculations instead of electrical calculations.

[0026] In this embodiment, for the matrix involved in the artificial intelligence task, the main eigenvalue of the matrix (that is, the eigenvalue with the largest absolute value) and the matrix operation required for the corresponding eigenvector can be recursively calculated according to the power iteration algorithm. If the matrix operation is matrix multiplication or matrix and vector multiplication, the matrix operation involved in the matrix iteration process can be implemented by the photonic processor 10. The photon processor 10 is constructed by optical silicon-based devices and may include at least an operation process simulation optical path, a loop iteration control optical path, and a result reading optical path; the operation process simulation optical path includes a plurality of tunable photonic devices, based on the spatial position of each photonic device corresponding to the position of the matrix element, the value of each element is determined by adjusting the phase of the corresponding photonic device, and each photonic device is deployed according to an iterative operation process of the matrix to be calculated; the loop iteration control optical path compensates for the transmission loss of the optical signal output by the operation process simulation optical path after completing an iterative operation, and controls the optical compensation signal corresponding to the first k-1 iterative operations to be transmitted to the operation process simulation optical path for the next iterative operation, and controls the optical compensation signal corresponding to the k-1th and kth iterative operations to be transmitted to the result reading optical path; the result reading optical path reads the first data signal corresponding to the k-1th iterative operation, and transmits the first data signal to the operation process simulation optical path for the kth iterative operation; reads the second data signal corresponding to the kth iterative operation, and the second data signal is the main eigenvalue of the matrix and its corresponding eigenvector. If other eigenvalues ​​need to be calculated, according to the eigenvalues ​​of the current matrix A and matrix A and its corresponding eigenvector , through the relationship A new matrix is ​​constructed. By dynamically adjusting the phase of each photonic device, the elements of matrix A are adjusted to the corresponding elements of the new matrix. This matrix is ​​used as the current matrix. The above process is executed using the photonic processor to obtain the second largest eigenvalue and its corresponding eigenvector. The cycle is repeated until all the eigenvalues ​​of the matrix and the corresponding eigenvectors are obtained.

[0027] From the above, it can be seen that when the photonic processor of this embodiment performs matrix operations involved in artificial intelligence tasks, each calculation process of the eigenvalue and its eigenvector only requires two photoelectric / electro-optical conversions, which can effectively save resources and improve computing power by using a photonic processor with fewer photoelectric / electro-optical conversions to calculate the matrix characteristic problem.

[0028] It should be noted that the above application scenarios are only shown to facilitate understanding of the ideas and principles of the present invention, and the embodiments of the present invention are not limited in this respect. On the contrary, the embodiments of the present invention can be applied to any applicable scenario. After introducing the technical solution of the present invention, various non-limiting embodiments of the present invention are described in detail below with reference to the accompanying drawings and specific embodiments. First, please refer to Figure 2 , Figure 2The following is a structural diagram of a photon processor provided in this embodiment in an exemplary implementation manner. This embodiment may include the following contents: The photonic processor includes a computational simulation optical circuit 201, a loop-iteration control optical circuit 202, and a result-reading optical circuit 203. These circuits are all constructed based on optical components. Leveraging the parallel propagation and optical interference properties of optical signals within the optical circuit, they replace electrons as information carriers for data processing, enabling efficient, full-domain parallel computation of matrices. The basic principle is to encode the initial input vector or matrix into multi-path parallel light intensity / phase signals. These signals are dynamically mapped to preset matrix weights via tunable photonic devices (such as Mach-Zehnder interferometer arrays). During transmission, the multipath light signals undergo weight loading and light field interference superposition, naturally performing multiplication-addition operations. Finally, at the output, for example, a grating coupler array can be used to convert the multidimensional interference light intensity distribution into electrical signals, and the computational results are read in parallel via photodetectors.

[0029] The computation process simulation optical circuit 201 includes three signal inputs: a first signal input for receiving a laser signal emitted by a laser, a second signal input for receiving a compensation optical signal from the previous iteration transmitted by the cyclic iteration control optical circuit 202, and a third signal input for receiving the k-1th read data from the result reading optical circuit 203. The output of the computation process simulation optical circuit 201 is connected to the input of the cyclic iteration control optical circuit 202, transmitting the output result of each iteration to the cyclic iteration control optical circuit 202. The cyclic iteration control optical circuit 202 includes two output channels, one of which is connected to the second signal input of the computation process simulation optical circuit 201, and the other is connected to the input of the result reading optical circuit 203. One output of the result reading optical circuit 203 is used to output the result, and the other output is connected to the third signal input of the computation process simulation optical circuit 201. Different photonic devices can be connected via optical waveguides, such as optical fibers or silicon optical waveguides.

[0030] In this embodiment, if Figure 3As shown, the computational process simulation module 201 includes multiple tunable photonic devices, each spatially positioned to correspond to a matrix element position. Each photonic device is deployed according to a single iterative computation of the matrix to be computed, where the deployment includes the position of each photonic device and the connections between them. The value of each element is obtained by adjusting the phase of the corresponding photonic device through thermo-optical modulation or electro-optical modulation. The matrix to be computed can obtain characteristic data after k iterative computations, where these k iterative computations are matrix multiplications or matrix-vector multiplications, and the characteristic data of the matrix to be computed is determined based on the output of the k-1 iteration and the output of the kth iteration. The cyclic iteration control optical circuit 202 includes at least one input and two output channels, namely, a first optical output channel and a second optical output channel. The first optical output channel of the cyclic iteration control optical circuit 202 is connected to the input of the computational process simulation optical circuit 201, and the second optical output channel is connected to the result reading optical circuit 203. The cyclic iteration control optical path 202 also includes an optical path switching control device that controls which output channel is selected. The optical path switching control device can be set manually or triggered by an automated program. The optical path switching control device transmits the optical compensation signals corresponding to the first k-1 iterative operations to the operation process simulation optical path 202 through the first optical path output channel for the next iterative operation, and transmits the optical compensation signals corresponding to the k-1th and kth iterative operations to the result reading optical path 203 through the second optical path output channel. Furthermore, considering that optical signals may experience transmission losses between different components, such as MZI couplers and optical waveguides between different components, in order to avoid transmission losses in subsequent cyclic iterations, the iterative operation optical signal output by the operation process simulation optical circuit 202 can also be compensated. For ease of description, the compensated signal is defined as an optical compensation signal. Accordingly, the cyclic iteration control optical circuit 202 compensates for the transmission losses of the optical signal output by the operation process simulation module 201 after completing an iterative operation, controls the transmission of the optical compensation signals corresponding to the first k-1 iterations to the operation process simulation module for the next iterative operation, and controls the transmission of the optical compensation signals corresponding to the k-1th and kth iterations to the result reading optical circuit 203. The result reading optical circuit 203 transmits the first data signal corresponding to the k-1th iterative operation to the operation process simulation optical circuit 201 for the kth iterative operation.In other words, when the operation process simulation optical circuit 202 completes the k-1th iterative operation, the optical signal output by the k-1th iterative operation is output to the loop iteration control optical circuit 202, and the loop iteration control optical circuit 202 transmits it to the result reading optical circuit 203 through the second output channel. The result reading optical circuit 203 reads the optical signal and stores it in the electronic storage unit. For the convenience of description, the signal output and compensated by the k-1th iterative operation and the signal obtained after reading processing are defined as the first data signal, and then the first data signal is transmitted to the operation process simulation module 201 for the kth iterative operation. After the operation process simulation module 201 completes the kth iterative operation, it outputs it to the result reading optical circuit 203 through the second optical path output channel. The result reading optical circuit 203 reads the signal to obtain a second data signal. The second data signal and the data corresponding to the first data signal are subjected to corresponding operations to obtain characteristic data, and the obtained characteristic data is output through the output end of the result reading optical circuit 203. Exemplarily, the result reading optical path 203 may include at least a plurality of photodetectors, that is, the final photodetector obtains an output to represent the main eigenvalue and the corresponding eigenvector of the matrix.

[0031] In the technical solution provided in this embodiment, for a matrix whose characteristic data is calculated through multiple cyclic iterative processes, based on the mapping relationship between one tunable photonic device and one matrix element, a group of tunable photonic devices can be used to simulate a single iterative process of the matrix, and the optical signal output of each iterative operation is compensated by controlling the optical path through cyclic iterative control, and the optical signal after the previous k-1 compensation is controlled to enter the calculation process simulation optical path according to the current number of iterations to participate in the next iterative operation, and the optical compensation signals of the k-1th and kth times are output from the optical domain to the electronic storage unit, and then the optical compensation signal of the k-1th time is converted back into an optical signal. The signal input loop iteration controls the optical path to participate in the last iteration. The entire process only requires 2 photoelectric / electro-optical conversions, which reduces k-2 photoelectric / electro-optical conversions compared with related technologies. It not only effectively avoids the increase in system delay caused by photoelectric signal conversion and improves the matrix data processing efficiency, but also avoids the energy consumption overhead caused by k-2 photoelectric / electro-optical conversions, saves resources, effectively reduces the power consumption of optical computing, effectively improves the computing power of the photonic processor, and avoids the problem of limited data transmission rate due to the sampling rate and bandwidth limitations of photoelectric / electro-optical conversion that cannot fully match the high-speed characteristics of optical computing, thereby further improving the matrix data processing efficiency.

[0032] The above embodiment does not impose any restrictions on how the computational process simulation optical path simulates a single iterative matrix operation. This embodiment also provides an exemplary implementation method. In this embodiment, the computational process simulation optical path can be implemented by repeatedly applying the matrix to be calculated to an arbitrary initial input array. The initial input array can be a vector or a matrix, so that the result after each iteration gradually converges to the direction of the principal eigenvector. Accordingly, the computational process simulation optical path 201 can include an initial signal input encoding array and a matrix representation optical device array. The initial signal input encoding array includes a plurality of tunable first-type photonic devices, each of which is arranged according to the dimensions and array element arrangement of the initial input array, such that each first-type photonic device corresponds one-to-one with an array element of the initial input array. Each first-type photonic device encodes the input optical emission signal into the corresponding array element value by adjusting the phase. The matrix representation optical device array includes a plurality of tunable second-type photonic devices, each of which is arranged according to the row dimensions, column dimensions, and matrix element arrangement of the matrix to be calculated, such that each second-type photonic device corresponds one-to-one with a matrix element of the matrix to be calculated. The connection relationship between each first-class photonic device and each second-class photonic device is determined according to the multiplication operation between the initial input array and the matrix to be calculated. Each first-class photonic device outputs the encoded optical coding signal to the corresponding second-class photonic device, and each second-class photonic device outputs the received optical coding signal to the loop iteration control optical path to complete an iterative operation of calculating the characteristic data of the matrix to be calculated.

[0033] In this embodiment, if Figure 3 As shown, the initial signal input encoding array is the initial input array To encode, each array element Corresponding to an independent first-class photonic device, , n is the array dimension, which is also the total number of first-class photonic devices. The phase of the first-class photonic devices is modulated to control the dynamic changes of the array elements. The laser emits an optical signal into each first-class photonic device. An external electrical signal can be used to change the phase difference of each first-class photonic device so that the value of the optical transmission rate function is equal to the value of the array vector element or matrix element, thereby converting the electrical signal into an optical signal input. The current methods for controlling the phase include electro-optical / thermo-optical modulation, that is, adjusting the input voltage or the temperature of the device through an electrical signal to achieve phase adjustment. The matrix represents the optical device array as a network array formed by the second-class photonic devices to represent the matrix A to be calculated. Each matrix element of the matrix A to be calculated corresponds to the optical power transmission function of a second-class photonic device. The dynamic control of the matrix elements is achieved by adjusting the phase difference. In this way, correspondingly, the output of the first-class photonic device represents the initial input array The optical signal enters the optical device array represented by the matrix A to be calculated. At this time, the matrix represents the optical signal output by the optical device array, which represents the output vector ,like Figure 5 shown.

[0034] The above embodiment does not impose any restrictions on the structures of the first type of photonic device and the second type of photonic device. It is sufficient that the first type of photonic device and the second type of photonic device are tunable photonic components, such as any type of interferometer. This embodiment also provides an exemplary structure of the first type of photonic device and the second type of photonic device, which may include the following: For the first type of photonic devices, such as Figure 4 As shown, the structures of the first-type photonic devices are the same. Each of the first-type photonic devices may include a first input coupler, a third phase shifter, a fourth phase shifter, and a first output coupler. The first input end of the first input coupler is connected to the laser, the output end of the first input coupler is connected to one end of the third phase shifter and the fourth phase shifter, the other ends of the third phase shifter and the fourth phase shifter are both connected to the input end of the first output coupler, and the third output end of the first output coupler is connected to the input end of the corresponding second-type photonic device. The first input coupler divides the optical emission signal input through the first input end into a first optical emission signal and a second optical emission signal. In order to facilitate Description, the signal received by the first type of photonic device is defined as a light emission signal, and the two light signals divided into the signal are defined as a first light emission signal and a second light emission signal. The first light emission signal is input into a third phase shifter, and the second light emission signal is input into a fourth phase shifter. The corresponding array element value is obtained by adjusting the phase difference between the light emission signal output by the third phase shifter and the light emission signal output by the fourth phase shifter. The first output coupler combines the light emission signal output by the third phase shifter and the light emission signal output by the fourth phase shifter into an optical coding signal. The combined optical signal is defined as an optical coding signal. Finally, the optical coding signal is output through the third output end.

[0035] Exemplarily, the adjustment process of the array element value may be as follows: for each array element of the initial input array, the difference between twice the value of the current array element and 1 is used as the phase difference cosine value; based on the fact that twice the phase difference between the light emission signal output by the third phase shifter and the light emission signal output by the fourth phase shifter is the same as the phase difference cosine value, the current phase difference between the light emission signal output by the third phase shifter and the light emission signal output by the fourth phase shifter is calculated; and the phase difference between the light emission signal output by the third phase shifter and the light emission signal output by the fourth phase shifter is adjusted to the current value by a phase modulation method. For example, the phase difference calculation relationship may be pre-stored: , represents the phase difference between the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter, and a represents the value of the current array element.

[0036] For the second type of photonic devices, such as Figure 5 and Figure 6 As shown, the structures of the second-type photonic devices are the same. The second-type photonic devices include a second input coupler, a fifth phase shifter, a sixth phase shifter, and a second output coupler. The third input end of the second input coupler is connected to the output end of the corresponding first-type photonic device, the output end of the second input coupler is connected to one end of the fifth phase shifter and the sixth phase shifter, the other ends of the fifth phase shifter and the sixth phase shifter are both connected to the input end of the second output coupler, and the fifth output end of the second output coupler is connected to the input end of the corresponding signal amplifier. The second input coupler divides the optically coded signal input through the third input end into a first optically coded signal and a second optically coded signal, and inputs the first optically coded signal into the fifth phase shifter and the second optically coded signal into the sixth phase shifter. The corresponding matrix element value is obtained by adjusting the phase difference between the optically coded signal output by the fifth phase shifter and the optically coded signal output by the sixth phase shifter. The second output coupler combines the optically coded signal output by the fifth phase shifter and the optically coded signal output by the sixth phase shifter into an optical operation signal, and outputs the optical operation signal through the fifth output end.

[0037] Exemplarily, the adjustment process of the matrix element value may be as follows: for each matrix element of the calculation matrix, the difference between twice the value of the current matrix element and 1 is used as the target phase difference value; based on the cosine value of twice the phase difference between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter being the same as the target phase difference value, the phase difference value between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter is calculated; and the phase difference between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter is adjusted to the phase difference value by a phase modulation method. For example, the phase difference calculation relationship may be pre-stored: , represents the phase difference between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter, and b represents the value of the current matrix element.

[0038] Based on the above embodiment, the result reading optical path 203 detects a first data signal, and the first data signal includes at least an intermediate vector and a vector modulus with the same dimension as the initial input array. In order to realize the k-th iterative operation, it is necessary to control the initial signal input coding array of the optical path 202 through cyclic iteration to convert the first data signal into a corresponding optical signal to participate in the k-th iterative operation. Correspondingly, the vector elements of the intermediate vector correspond one-to-one to each first-type photonic device of the initial signal input coding array, and the phase difference of the corresponding first-type photonic device is adjusted according to the vector element value of the intermediate vector. For example, the phase value of each first-type photonic device is adjusted by electro-optical modulation or thermo-optical modulation so that the optical power transfer function value of the first-type photonic device is the same as the vector element value corresponding to the intermediate vector; in this way, each first-type photonic device inputs the adjusted optical signal to the corresponding second-type photonic device, and when each second-type photonic device outputs the corresponding optical signal, the k-th iterative operation is completed.

[0039] As can be seen from the above, this embodiment realizes the simulation of matrix operations through the above-mentioned first type of photonic devices and second type of photonic devices. The function can be switched by adjusting the phase in real time without hardware changes. It is more reconfigurable and versatile. It does not require energy to maintain the optical path itself. It only consumes energy during phase adjustment, which reduces energy consumption. The optical signal is not affected by circuit noise, is suitable for complex electromagnetic environments, and is conducive to improving data processing accuracy.

[0040] The above embodiments do not impose any limitation on the structure of the cyclic iterative control optical path. Based on the above embodiments, the present invention further provides an exemplary structure of the cyclic iterative control optical path, which may include the following contents: the cyclic iterative control optical path includes a signal amplifier group and an optical path selector group; the total number of signal amplifiers in the signal amplifier group is the same as the total number of optical path selectors in the optical path selector group; the signal amplifier group includes multiple signal amplifiers, the number of signal amplifiers is the same as the number of optical signal output paths of one iterative operation, one end of each signal amplifier is connected to a photonic device corresponding to an output optical signal of the initial signal input coding array, and the other end is connected to an input end of an optical path selector of the optical path selector group; each optical path selector of the optical path selector group includes a first optical path output channel, a second optical path output channel and an optical path switching control device, the first optical path output channel is connected to the operation process simulation optical path, the second optical path output channel is connected to the result reading optical path, and the optical path switching control device determines whether the optical compensation signal output by the corresponding signal amplifier enters the first optical path output channel or the second optical path output channel according to the current number of iterative operations.

[0041] In this embodiment, the signal amplifiers and optical path selectors of the iterative control optical path correspond one-to-one. Each signal amplifier and optical path selector corresponds to one optical signal output by the iterative control optical path. One optical signal represents the dimension of the output data obtained by a single iterative operation. For example, if the dimension of the initial input array is 4×1 and the dimension of the matrix to be calculated is 4×4, then the dimension of the output result is 4×1. Accordingly, if four optical signals are output, the number of signal amplifiers and optical path selectors in the iterative control optical path is four, respectively. In other words, the number of signal amplifiers and optical path selectors in the iterative control optical path corresponds to the dimension of the matrix or vector corresponding to the optical signal output by the iterative control optical path after completing one iterative operation.

[0042] Considering that when an optical signal passes through an optical waveguide, coupler, or MZI (Mach–Zehnder Interferometer), transmission loss may occur due to material absorption and scattering, insertion loss, or splitting loss. If the loss is too great, the signal strength will fall below the detection threshold at the receiving end, resulting in an increased bit error rate or communication interruption. This embodiment uses a signal amplifier to directly amplify the optical signal to compensate for the loss and maintain system performance. The signal amplifier can be any photonic device capable of transmitting and compensating for the optical signal. For example, an erbium-doped fiber amplifier (EDFA) can be used to compensate for the transmission loss of the optical waveguide, coupler, and interferometer that outputs the optical signal.

[0043] Furthermore, the present invention also provides an exemplary result of the optical path selector, such as Figure 7 As shown, the optical path selector may include a beam splitter, a first phase shifter, a second phase shifter and a beam combiner; the input end of the beam splitter is connected to the corresponding signal amplifier, and the output end of the beam splitter is connected to one end of the first phase shifter and the second phase shifter. The beam splitter divides the optical compensation signal received through the input end into a first optical compensation signal and a second optical compensation signal, and inputs the first optical compensation signal to the first phase shifter and the second optical compensation signal to the second phase shifter; the other ends of the first phase shifter and the second phase shifter are both connected to one end of the beam combiner, and the beam combiner includes a first output end and a second output end, the first output end is connected to the operation process simulation optical path, and the second output end is connected to the result reading optical path; the beam combiner combines the optical signals output by the first phase shifter and the second phase shifter into a new optical signal, and outputs the new optical signal through the first output end or the second output end.

[0044] In this embodiment, the optical path selector implements cyclic iterative operations by controlling whether the optical signal output from an iterative operation enters the operation process simulation optical path or is read by the result reading optical path. When the optical path selector is configured in the "On" state, the output optical signal is allowed to enter the operation process simulation optical path. When configured in the "Off" state, the optical signal is allowed to enter the photodetector for result reading. The transmission matrix corresponding to the above optical path selector can be expressed as: .

[0045] in, , i represents the imaginary unit. When the input The port inputs the compensation optical signal, and the other input does not input the signal, that is, , then we can calculate from Input to output or The output optical power transfer function is: Through electro-optical / thermo-optical modulation, that is, adjusting the input voltage or device temperature through electrical signals, the phase difference is achieved. When When it is on, it indicates the “On” state, and the output signal is The output enters the feedback loop, when When it is in the "Off" state, the output signal is The output enters the photodetector to obtain the result. That is, for example, for the first k-1 iterations, the phase difference between the optical compensation signal output by the first phase shifter and the optical compensation signal output by the second phase shifter is adjusted to , so that the new optical signal is input into the calculation process simulation optical path through the first output end; for the k-1th and kth iterative operations, the phase difference between the optical compensation signal output by the first phase shifter and the optical compensation signal output by the second phase shifter is adjusted to 0, so that the new optical signal is input into the result reading optical path through the second output end.

[0046] As can be seen from the above, this embodiment compensates for optical signal transmission loss by iteratively controlling the optical path, is not limited by transmission distance, and can achieve multiple iterations. Optical path selection is achieved through phase adjustment, which not only has a fast switching speed but also extremely low energy consumption.

[0047] Considering that the characteristic problem calculation process of the matrix to be calculated is sometimes not achievable using a photon processor, or the overall cost is higher than that of an electronic processor, the present invention can also recursively summarize the results of k iterations of the matrix to be calculated using a power iteration algorithm and extract the corresponding matrix operation. If the matrix operation process is a matrix multiplication operation or a matrix and vector multiplication operation, the photon processor of the present invention is used to process the matrix to be calculated, which may include the following: The calculation process simulates the optical path including an initial signal input coding array and a matrix representation optical device array; the initial signal input coding array includes a group of tunable first-type photonic devices arranged in a longitudinal direction and the same number as the dimension of the initial input vector, each first-type photonic device corresponds one-to-one to the vector element of the initial input vector, and the input optical signal is encoded into the corresponding vector element value by adjusting the phase; the matrix representation optical device array includes multiple tunable second-type photonic devices, each second-type photonic device is arranged according to the row dimension, column dimension and matrix element arrangement of the matrix to be calculated, so that each second-type photonic device corresponds one-to-one to the matrix element of the matrix to be calculated; the connection relationship between each first-type photonic device and each second-type photonic device is determined according to the multiplication operation method between the initial input vector and the matrix to be calculated, each first-type photonic device outputs the encoded optical signal to the corresponding second-type photonic device, and each second-type photonic device outputs the received optical signal to the cyclic iteration control optical path to complete an iterative operation of calculating the main eigenvalue and the corresponding main eigenvector of the matrix to be calculated through the power iteration algorithm.

[0048] In this embodiment, the matrix to be calculated is subjected to k matrix-vector multiplication operations or matrix multiplication operations through a power iteration algorithm to obtain a principal eigenvalue and a corresponding principal eigenvector. The process of inferring the principal eigenvalue (i.e., the eigenvalue with the largest absolute value) of a real symmetric matrix such as the matrix to be calculated and its corresponding eigenvector through the power iteration algorithm may include: Iteration 0: Select the initial input vector , and normalize the initial input vector to a unit vector .

[0049] Iteration 1: , The vector output by the first iteration is normalized to a unit vector .

[0050] Iteration 2: , The vector output by the second iteration is based on an arbitrary constant Multiplying a vector The overall modulus is equal to a constant times The absolute value of the vector multiplied by The modulus length, that is , the second iteration of the relationship Defined as , then the vector Normalized to unit vector .

[0051] 3rd iteration: , The vector output of the third iteration is Normalized to unit vector Continuously recursively, the result of the k-th iteration operation is: .

[0052] Preset accuracy ,like When , the power iteration algorithm stops iterating and outputs the main eigenvalue , the corresponding eigenvector is The number of iterations and the accuracy of the power iteration algorithm are determined by the matrix characteristics and accuracy, and their values ​​can be adjusted according to the actual application. According to the results of the k-th iteration operation, the result of the k-th iteration is It is necessary to perform k matrix-vector multiplication operations, and the main eigenvalue output requires the results of the k-1th and kth iterations. Accordingly, this embodiment modulates the input optical signal by inputting the initial signal into the coding array, and outputs the optical signal representing the initial input vector to the matrix representation optical device array, such as Figure 8 As shown, the matrix represents the optical device array and the matrix to be calculated is represented. When the optical signal passes through the matrix and is output from the optical device array, an iterative operation process is completed.

[0053] Since the power iteration algorithm is only used to determine the main eigenvalue of the matrix to be calculated (i.e., the eigenvalue with the largest absolute value) and its corresponding eigenvector In order to determine other eigenvalues ​​and eigenvectors, this embodiment also adopts the following implementation method: According to the matrix to be calculated, the principal eigenvalue and the principal eigenvector, a target matrix is ​​determined; the target matrix is ​​iterated k times by a power iteration algorithm to obtain the second largest eigenvalue and the corresponding second eigenvector of the matrix to be calculated; the matrix elements of the target matrix correspond one-to-one to each second type of photonic device in the matrix representation optical device array, and the phase difference of the corresponding second type of photonic device is adjusted according to the matrix element value of the target matrix so that the optical power transfer function value of the second type of photonic device is the same as the corresponding matrix element value of the target matrix; the laser is controlled to emit an optical signal again and input it to the initial signal input coding array, and after the operation process simulates the optical path to complete k iterative operations, the result of the kth iterative operation read by the optical path is read as the second largest eigenvalue and the corresponding second eigenvector.

[0054] In this embodiment, the relationship To obtain the target matrix, the target matrix is ​​used to replace the matrix in the previous round of eigenvalue calculation process. For example, when calculating the second largest eigenvalue, the target matrix replaces the matrix to be calculated in the above embodiment, and the process of calculating the main eigenvalue and the corresponding eigenvector by the above photon processor can obtain the second largest eigenvalue of the matrix to be calculated. and the corresponding eigenvectors When calculating the third largest eigenvalue, the target matrix replaces the target matrix of the above embodiment, and the process of calculating the second largest eigenvalue and the corresponding eigenvector of the above photon processor is used to obtain the third largest eigenvalue and the corresponding eigenvector of the matrix to be calculated, and so on until the matrix All eigenvalues ​​and corresponding eigenvectors are extracted. The method of using the target matrix to replace the matrix in the last eigenvalue calculation process can be: the target matrix is ​​A1, and the matrix is ​​adjusted according to the values ​​of each matrix element of A1 to represent the phase of each second-class photonic device in the optical device array, so as to represent the real symmetric matrix A1, that is, each second-class photonic device corresponds to each matrix element in the second-class photonic device of the matrix, and the phase difference is adjusted by thermo-optical / electro-optical effect to realize the dynamic control of the matrix element. The above-mentioned photonic processor can be used to obtain The main eigenvalue of , the eigenvector is , which is the second largest eigenvalue of the matrix to be calculated and its corresponding eigenvector. Continue to construct , then, the above-mentioned photon processor can be used to obtain The main eigenvalue of , the eigenvector is , that is, the third largest eigenvalue of the matrix to be calculated and its corresponding eigenvector are obtained. By continuing in this way, all the eigenvalues ​​and corresponding eigenvectors of the matrix to be calculated can be obtained.

[0055] Finally, in order to make those skilled in the art more clearly understand the implementation of the present invention, the present invention also provides a schematic photon processor. Figure 8 The architecture of the photonic processor shown is taken as an example. Figure 8Taking a 4×4 real symmetric matrix as an example, the matrix is ​​represented by 16 second-class photonic devices cascaded into a network array. The tunable photonic devices in the computational simulation optical path of this embodiment are all implemented using silicon-based MZIs (Mach–Zehnder interferometers). Each MZI in the computational simulation optical path controls the input signal at one input port, the other input port to be silent, and outputs the signal through one output port. The optical path selector for iteratively controlling the optical path is also implemented using a silicon-based MZI. The optical signal compensation component can be an erbium-doped fiber amplifier, and the result reading optical path is a set of photoelectric converters. Each silicon-based MZI can be composed of two couplers and two phase shifters. They achieve phase-differential interference by adjusting the optical path difference. When the input optical signal passes through the first coupler, it is split into two beams, which pass through two different phase shifters to generate a certain phase difference. These two beams are then combined into a new optical signal through the second coupler, forming an interference signal. By adjusting the phase difference, the amplitude and phase of the optical signal can be precisely controlled, thereby completing the corresponding matrix operation.

[0056] In this embodiment, the matrix to be calculated is A real symmetric matrix of dimension , whose eigenvalue is , we can first use the power iteration algorithm to recursively calculate the main eigenvalue of the matrix to be calculated and the matrix operation of the corresponding eigenvector. When it is determined that the matrix operation of the matrix to be calculated is a matrix multiplication operation or a matrix-vector multiplication operation, it is determined that the matrix calculation process of the matrix to be calculated can be implemented using the photon processor of this embodiment. The process of recursively calculating the matrix A to be calculated by the power iteration algorithm is: the number of iteration steps , first select an initial vector and normalize it, that is, select the initial input vector , and normalize it to a unit vector , , Is the modulus of the vector. Then in each iteration, the vector is multiplied by the matrix and the result is normalized again, that is, the calculation , and normalize it to a unit vector , , loop until the vector changes below the preset accuracy The result of the kth iteration can be expressed as ,when When , the algorithm stops iterating and outputs the main eigenvalue , the corresponding eigenvector is , recursively calculate the result of the kth iteration according to the above algorithm steps and extract the required matrix operations. Therefore, when using the photonic processor of this embodiment to calculate the eigenvalues ​​of the matrix to be calculated, it is only necessary to read the results of the k-1th and kth iterations.

[0057] The process of calculating the main eigenvalue of the matrix A to be calculated by the photon processor may include the following: A1: The laser emits an optical signal into each silicon-based MZI, and the output represents the initial input vector by adjusting the phase The optical signal will represent the initial input vector The optical signal input represents the real symmetric matrix A, and the matrix composed of the silicon-based MZIs represents the optical device array. At this time, the output optical signal completes a matrix-vector multiplication operation, that is, .

[0058] A2: The optical signal output by the matrix represents the optical device array. The optical signal is amplified by the erbium-doped fiber amplifier to compensate for the transmission loss of the optical waveguide, coupler and MZI. The optical signal is fed back to the matrix of the optical device array composed of silicon-based MZI units representing the real symmetric matrix through the MZI loop that selects the optical path to complete the matrix-vector multiplication operation, that is, , and will represent The optical signal is again input to the matrix representation optical device array.

[0059] A3: Repeat step A2 until the k-1th iteration completes the acquisition .

[0060] A4: Will represent The optical signal is transmitted through the MZI switch to the photodetector for acquisition. Information, The information includes vector The element value and the modulus of the vector The laser emits a signal again, and the output represents the The light signal will represent The optical signal input represents a real symmetric matrix of each silicon-based MZI, which is composed of a matrix representing the optical device array, completing the last matrix-vector multiplication operation, that is, .

[0061] A5: Will represent The optical signal is transmitted through the MZI switch to the photodetector for acquisition. Finally, according to , the corresponding eigenvector is , we can get the main eigenvalue and the corresponding eigenvector.

[0062] From the above, it can be seen that this embodiment first recursively calculates the matrix main eigenvalues ​​and the matrix operations required for the corresponding eigenvectors based on the power iteration algorithm, and then designs a photon processor with fewer photoelectric / electro-optical conversion times to implement the matrix operations involved in the above iterative process to complete the calculation tasks of the main eigenvalues ​​and eigenvectors. Finally, a specific matrix is ​​constructed and the above-mentioned photon processor is used to sequentially calculate all the eigenvalues ​​and corresponding eigenvectors of the original matrix. The matrix characteristic problem can be processed by a photon processor with fewer photoelectric / electro-optical conversion times, effectively improving data processing efficiency and saving resources.

[0063] Finally, the present invention also provides an implementation process of using the photon processor described in the above embodiment to execute an actual task that can be performed by calculating the eigenvalue of the matrix as the task execution result, see Figure 9 , Figure 9 This is a flow chart of a target data identification method provided in this embodiment. This embodiment may include the following contents: S901: When a data identification task of identifying target data from task data is received, task data related to the data identification task is acquired.

[0064] Data identification tasks are tasks that can be solved by calculating matrix eigenvalues, including but not limited to data dimensionality reduction, pattern recognition, network analysis, and model optimization. For example, principal component analysis achieves data dimensionality reduction by solving the eigenvalue problem of the covariance matrix (the matrix to be calculated). This means that the principal component analysis algorithm solves the eigenvalue problem of the covariance matrix to transform data from high-dimensional to low-dimensional data. Spectral clustering algorithms rely on the eigenvalues ​​of the Laplace matrix (the matrix to be calculated) to classify data. The PageRank (link analysis) algorithm uses the eigenvectors corresponding to the maximum eigenvalues ​​of the adjacency matrix (the matrix to be calculated) to rank web pages by importance. Task data is the data required for the analysis of data identification tasks, and target data is the data in the task data corresponding to the eigenvalues.

[0065] S902: Represent the task data in matrix form to obtain a matrix to be calculated, where characteristic data of the matrix to be calculated is the target data.

[0066] Among them, the task data can be constructed using any matrix form construction method such as covariance matrix, Laplace matrix, adjacency matrix, etc. Correspondingly, the matrix to be calculated is covariance matrix, Laplace matrix, adjacency matrix.

[0067] S903: When it is determined that the characteristic data calculation process of the matrix to be calculated includes multiple iterative matrix-vector multiplication operations or multiple iterative matrix multiplication operations, the matrix to be calculated is processed by the photon processor to obtain the characteristic data of the matrix to be calculated as the processing result of the data recognition task.

[0068] The target data can be the data corresponding to the main eigenvalue or the data corresponding to multiple eigenvalues. The final processing result is determined according to different tasks. If the data identification task is a hot event identification task, the processing result of the data identification task is the data corresponding to the main eigenvalue of the matrix to be calculated. If the data identification task is a data importance sorting task, the processing result of the data identification task is the sorting result of the data corresponding to multiple eigenvalues. For example, each eigenvalue of the matrix to be calculated is obtained, and the absolute value of each eigenvalue is calculated; the eigenvalues ​​are sorted in order of absolute value from large to small or from small to large to obtain feature sorting information; the absolute value of each eigenvalue is proportional to the importance of the corresponding data; the data corresponding to each eigenvalue is sorted according to the feature sorting information to obtain the sorting result corresponding to the data importance sorting task.

[0069] In the technical solution provided in this embodiment, the matrix data processing efficiency is effectively improved through the photonic processor, thereby effectively improving the target data recognition efficiency, improving the task response performance, saving the resources required in the execution of the target data recognition task, and being able to be deployed on resource-constrained edge devices with better practicality.

[0070] To help those skilled in the art better understand the implementation of the present invention, the present invention also provides an exemplary task implementation method. The data recognition task in this embodiment is to detect hot events on a social platform. The process of detecting hot events may include the following: B1: Construct the adjacency matrix corresponding to the dynamic graph based on the dynamic social network structure , t is the time.

[0071] In this embodiment, a hot event is a concentrated interaction of a large number of users around a specific topic in a short period of time, such as an emergency. The graph structure of the dynamic social network (user follow-up relationship, tweet forwarding) will change rapidly with the outbreak of the event. This step will construct the matrix to be calculated based on the dynamic social network structure. The adjacency matrix corresponding to the hot event A sudden change in the main eigenvalue, such as a sharp increase in the maximum eigenvalue, reflects a surge in network activity, thereby combining dynamic graph updates with fast feature extraction of hot events to solve the problem that traditional static graph models cannot capture such mutations in real time.

[0072] Exemplarily, the process of constructing an adjacency matrix may include: obtaining each entity node, each preset time step, and the corresponding preset time interval of the dynamic social network, and assigning a unique index value to each entity node; constructing an initial matrix of matrix element values ​​to be filled based on the total number of entity nodes, wherein the row dimension and column dimension of the initial matrix are both the same as the total number of entity nodes; within each preset time step, in each preset time interval of the current preset time step, determining the corresponding matrix element value of the initial matrix based on whether different entity nodes have a relationship, and using the initial matrix filled with the matrix element values ​​as the matrix to be calculated.

[0073] Among them, entity nodes refer to entities with dynamic attributes, such as users. The total number of nodes in the entire dynamic social network is N, and each node is assigned an index, such as 0 to N-1. For each time step t, create an N×N matrix and initialize it to all 0s, such as Then, for each edge that is active in time interval t, which represents the relationship between different entity nodes, such as user follow-up relationship, interaction or forwarding between users, the corresponding position in the adjacency matrix is ​​set to 1. In this way, an adjacency matrix is ​​obtained for each time step.

[0074] For example, suppose there are three entity nodes {v0, v1, v2}, and the preset time intervals are divided into two: [0, 5) and [5, 10]. Entity nodes v0 and v1 interacted in the time interval [0, 5), entity nodes v1 and v2 interacted in the time interval [5, 10], and entity nodes v1 and v0 interacted in the time intervals [0, 5) and [5, 10]. Correspondingly, the following adjacency matrix can be obtained for the two intervals: .

[0075] B2: Using the above photonic processor to quickly obtain the adjacency matrix The main eigenvalue of The main eigenvalue information is used to detect hot events.

[0076] It should be noted that there is no strict order in which the steps in the present invention are performed. As long as they conform to a logical order, the steps can be performed simultaneously or in a predetermined order. Figure 9 This is just a schematic and does not mean that this is the only execution order.

[0077] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method.

[0078] The present invention also provides a corresponding device for the target data identification method, which further makes the method more practical. Among them, the device can be described from the perspective of functional optical path and hardware. The target data identification device provided by the present invention is introduced below. The device is used to implement the target data identification method provided by the present invention. In this embodiment, the target data identification device may include or be divided into one or more program optical paths. The one or more program optical paths are stored in a storage medium and executed by one or more processors to complete the target data identification method disclosed in Example 1. The program optical path referred to in this embodiment refers to a series of computer program instruction segments that can complete specific functions, which is more suitable for describing the execution process of the target data identification device in the storage medium than the program itself. The following description will specifically introduce the functions of each program optical path in this embodiment. The target data identification device described below and the target data identification method described above can be referenced to each other.

[0079] Based on the angle of the functional optical path, see Figure 10 , Figure 10 This is a structural diagram of a target data identification device provided in this embodiment under a specific implementation mode. The device may include: The task data acquisition module 101 is configured to acquire task data related to the data identification task upon receiving a data identification task of identifying target data from task data.

[0080] The data processing module 102 is used to express the task data in a matrix form to obtain a matrix to be calculated, and the characteristic data of the matrix to be calculated is the target data.

[0081] The task execution module 103 is used to process the matrix to be calculated through a photon processor when it is determined that the characteristic data calculation process of the matrix to be calculated includes multiple iterative matrix-vector multiplication operations or multiple iterative matrix multiplication operations, so as to obtain the characteristic data of the matrix to be calculated as the processing result of the data recognition task.

[0082] Illustratively, in some implementations of this embodiment, the task execution module 103 is used to perform a hot event identification task, and uses the data corresponding to the main eigenvalue of the matrix to be calculated as the processing result of the data identification task.

[0083] As an exemplary implementation of the above embodiment, the above data processing module 102 can also be used to: obtain each entity node, each preset time step and the corresponding preset time interval of the dynamic social network, and assign a unique index value to each entity node; construct an initial matrix of matrix element values ​​to be filled based on the total number of entity nodes, and the row dimension and column dimension of the initial matrix are the same as the total number of entity nodes; within each preset time step, in each preset time interval of the current preset time step, determine the corresponding matrix element value of the initial matrix based on whether different entity nodes have a relationship, and use the initial matrix filled with matrix element values ​​as the matrix to be calculated.

[0084] Illustratively, in some other implementations of this embodiment, the above-mentioned task execution module 103 is used for the data importance sorting task, obtains each eigenvalue of the matrix to be calculated, and calculates the absolute value of each eigenvalue; sorts each eigenvalue in order from large to small or from small to large in absolute value to obtain feature sorting information; the absolute value of each eigenvalue is proportional to the importance of the corresponding data; the data corresponding to each eigenvalue is sorted according to the feature sorting information to obtain the sorting result corresponding to the data importance sorting task.

[0085] The target data recognition device mentioned above is described from the perspective of a functional optical path. Furthermore, the present invention also provides an electronic device, which is described from the perspective of hardware. Figure 11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present invention in one implementation manner. The electronic device includes a memory 111 and a processor 112. The memory 111 stores a computer program, and the processor 112 is configured to run the computer program to perform the steps of any of the above-mentioned target data recognition method embodiments.

[0086] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above-mentioned target data identification method embodiments when run.

[0087] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0088] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any of the above-mentioned target data identification method embodiments are implemented.

[0089] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above-mentioned target data identification method embodiments are implemented.

[0090] The above is a detailed introduction to a photon processor and target data identification method, device, electronic device, computer-readable storage medium, and computer program product provided by the present invention. The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. Whether the units and algorithm steps of each example described in each disclosed embodiment are executed in electronic hardware or computer software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, and such implementation should not be considered to exceed the scope of the present invention. Without departing from the principles of the present invention, the present invention can also be improved and modified in a number of ways, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. A photonic processor, characterized in that: Including calculation process simulation optical path, loop iteration control optical path and result reading optical path; The calculation process simulation optical path includes a plurality of tunable photonic devices deployed according to a single iterative calculation process of the matrix to be calculated, and the output end of the tunable photonic devices is connected to the input end of the cyclic iterative control optical path; the matrix to be calculated is iteratively calculated k times to obtain characteristic data; The first optical path output channel of the cyclic iteration control optical path is connected to the input end of the calculation process simulation optical path, and the second optical path output channel is connected to the result reading optical path. The optical path switching control device transmits the optical compensation signal corresponding to the first k-1 iteration calculations to the calculation process simulation optical path for the next iteration calculation through the first optical path output channel, and transmits the optical compensation signal corresponding to the k-1th and kth iteration calculations to the result reading optical path through the second optical path output channel; The result reading optical path is connected to the input end of the operation process simulation optical path, and the first data signal corresponding to the k-1th iterative operation is transmitted to the operation process simulation optical path for the kth iterative operation, and the characteristic data is obtained based on the first data signal and the second data signal corresponding to the kth iterative operation.

2. The photonic processor according to claim 1, wherein: The cyclic iterative control optical path includes a signal amplifier group and an optical path selector group; the total number of signal amplifiers in the signal amplifier group is the same as the total number of optical path selectors in the optical path selector group; The signal amplifier group includes a plurality of signal amplifiers, the number of which is the same as the number of optical signal output paths of one iterative operation, one end of each signal amplifier is connected to a photonic device corresponding to one optical signal outputted by the cyclic iterative control optical path, and the other end is connected to an input end of an optical path selector of the optical path selector group; Each optical path selector in the optical path selector group includes a first optical path output channel, a second optical path output channel and an optical path switching control device. The first optical path output channel is connected to the operation process simulation optical path, and the second optical path output channel is connected to the result reading optical path. The optical path switching control device determines whether the optical compensation signal output by the corresponding signal amplifier enters the first optical path output channel or the second optical path output channel based on the current number of iterative operations.

3. The photonic processor according to claim 2, wherein: The optical path selector includes a beam splitter, a first phase shifter, a second phase shifter and a beam combiner; The input end of the beam splitter is connected to the corresponding signal amplifier, and the output end of the beam splitter is connected to one end of the first phase shifter and the second phase shifter. The beam splitter splits the optical compensation signal received through the input end into a first optical compensation signal and a second optical compensation signal, and inputs the first optical compensation signal to the first phase shifter and the second optical compensation signal to the second phase shifter. The other ends of the first phase shifter and the second phase shifter are both connected to one end of the beam combiner. The beam combiner includes a first output end and a second output end. The first output end is connected to the calculation process simulation optical path, and the second output end is connected to the result reading optical path. The beam combiner combines the optical signals output by the first phase shifter and the second phase shifter into a new optical signal, and outputs the new optical signal through the first output end or the second output end.

4. The photonic processor according to claim 3, wherein: Outputting the new optical signal through the first output end or the second output end includes: For the first k-1 iterative operations, adjusting the phase difference between the optical compensation signal output by the first phase shifter and the optical compensation signal output by the second phase shifter to π, so that the new optical signal is input into the simulation optical path of the operation process through the first output end; For the k-1th and kth iterative operations, the phase difference between the optical compensation signal output by the first phase shifter and the optical compensation signal output by the second phase shifter is adjusted to 0, so that the new optical signal is input into the result reading optical path through the second output end.

5. The photonic processor according to claim 1, wherein: The operation process simulates an optical path including an initial signal input coding array and a matrix representation optical device array; The initial signal input encoding array includes a plurality of tunable first-type photonic devices, each of which is arranged according to the dimensions and array element arrangement of the initial input array, so that each first-type photonic device corresponds one-to-one to an array element of the initial input array, and each first-type photonic device encodes the input light emission signal into a corresponding array element value by adjusting the phase; The matrix represents that the optical device array includes a plurality of tunable second-type photonic devices, and each second-type photonic device is arranged according to the row dimension, column dimension and matrix element arrangement of the matrix to be calculated, so that each second-type photonic device corresponds one-to-one to the matrix element of the matrix to be calculated; The connection relationship between each first-class photonic device and each second-class photonic device is determined according to the multiplication operation method between the initial input array and the matrix to be calculated. Each first-class photonic device outputs the encoded optical coding signal to the corresponding second-class photonic device, and each second-class photonic device outputs the received optical coding signal to the loop iteration control optical path to complete an iterative operation of calculating the characteristic data of the matrix to be calculated.

6. The photonic processor according to claim 5, wherein: The first type of photonic device includes a first input coupler, a third phase shifter, a fourth phase shifter and a first output coupler; The first input end of the first input coupler is connected to a laser, the output end of the first input coupler is connected to one end of the third phase shifter and the fourth phase shifter, the other ends of the third phase shifter and the fourth phase shifter are both connected to the input end of the first output coupler, and the third output end of the first output coupler is connected to the input end of the corresponding second-type photonic device; The first input coupler divides the optical transmit signal input through the first input end into a first optical transmit signal and a second optical transmit signal, and inputs the first optical transmit signal into the third phase shifter, and inputs the second optical transmit signal into the fourth phase shifter. The corresponding array element value is obtained by adjusting the phase difference between the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter. The first output coupler combines the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter into an optically coded signal, and outputs the optically coded signal through the third output end.

7. The photonic processor according to claim 6, wherein: The method obtains a corresponding array element value by adjusting a phase difference between the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter, comprising: For each array element of the initial input array, the difference between twice the value of the current array element and 1 is used as the phase difference cosine value; calculating a current value of the phase difference between the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter based on the fact that twice the phase difference between the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter is the same as the cosine value of the phase difference; The phase difference between the optical transmit signal output by the third phase shifter and the optical transmit signal output by the fourth phase shifter is adjusted to the current value by a phase modulation method.

8. The photonic processor according to claim 5, wherein: The second type of photonic device includes a second input coupler, a fifth phase shifter, a sixth phase shifter and a second output coupler; The third input end of the second input coupler is connected to the output end of the corresponding first-type photonic device, the output end of the second input coupler is connected to one end of the fifth phase shifter and the sixth phase shifter, the other ends of the fifth phase shifter and the sixth phase shifter are both connected to the input end of the second output coupler, and the fifth output end of the second output coupler is connected to the input end of the corresponding signal amplifier; The second input coupler divides the optical coding signal input through the third input end into a first optical coding signal and a second optical coding signal, and inputs the first optical coding signal into the fifth phase shifter, and inputs the second optical coding signal into the sixth phase shifter. The corresponding matrix element value is obtained by adjusting the phase difference between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter. The second output coupler combines the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter into an optical calculation signal, and outputs the optical calculation signal through the fifth output end.

9. The photonic processor according to claim 8, characterized in that The corresponding matrix element value is obtained by adjusting the phase difference between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter, including: For each matrix element of the matrix to be calculated, the difference between twice the value of the current matrix element and 1 is used as the target phase difference; Calculating a phase difference between the optically coded signal output by the fifth phase shifter and the optically coded signal output by the sixth phase shifter based on the cosine value of twice the phase difference between the optically coded signal output by the fifth phase shifter and the optically coded signal output by the sixth phase shifter being the same as the target phase difference; The phase difference between the optical coding signal output by the fifth phase shifter and the optical coding signal output by the sixth phase shifter is adjusted to the phase difference value by a phase modulation method.

10. The photonic processor according to claim 5, wherein: The first data signal includes an intermediate vector and a vector modulus having the same dimension as the initial input array, and transmitting the first data signal to the operation process simulation optical path includes: The vector elements of the intermediate vector correspond one-to-one to the first-type photonic devices of the initial signal input coding array, and the phase differences of the corresponding first-type photonic devices are adjusted according to the vector element values ​​of the intermediate vector so that the optical power transfer function values ​​of the first-type photonic devices are the same as the vector element values ​​corresponding to the intermediate vector; Each first-type photonic device inputs the adjusted optical signal to the corresponding second-type photonic device. When each second-type photonic device outputs the corresponding optical signal, the k-th iterative operation is completed.

11. The photonic processor according to any one of claims 1 to 9, characterized in that: The matrix to be calculated is subjected to k matrix-vector multiplication operations or matrix multiplication operations through a power iteration algorithm to obtain a principal eigenvalue and a corresponding principal eigenvector, and the operation process simulates an optical path including an initial signal input coding array and a matrix representation optical device array; The initial signal input encoding array includes a group of tunable first-type photonic devices arranged in a longitudinal direction, the number of which is equal to the dimension of the initial input vector, each first-type photonic device corresponds one-to-one to a vector element of the initial input vector, and encodes the input optical signal into the corresponding vector element value by adjusting the phase; The matrix represents that the optical device array includes a plurality of tunable second-type photonic devices, and each second-type photonic device is arranged according to the row dimension, column dimension and matrix element arrangement of the matrix to be calculated, so that each second-type photonic device corresponds one-to-one to the matrix element of the matrix to be calculated; The connection relationship between each first-class photonic device and each second-class photonic device is determined according to the multiplication operation between the initial input vector and the matrix to be calculated. Each first-class photonic device outputs the encoded optical signal to the corresponding second-class photonic device, and each second-class photonic device outputs the received optical signal to the cyclic iteration control optical path to complete an iterative operation of calculating the main eigenvalue and the corresponding main eigenvector of the matrix to be calculated through the power iteration algorithm.

12. The photonic processor according to claim 11, wherein: Also includes: Determining a target matrix according to the matrix to be calculated, the main eigenvalue, and the main eigenvector; The target matrix is ​​iterated k times by a power iteration algorithm to obtain the second largest eigenvalue and the corresponding second eigenvector of the matrix to be calculated; The matrix elements of the target matrix correspond one-to-one to each second type of photonic device of the optical device array represented by the matrix, and the phase difference of the corresponding second type of photonic device is adjusted according to the matrix element value of the target matrix so that the optical power transfer function value of the second type of photonic device is the same as the matrix element value corresponding to the target matrix; The laser is controlled to emit an optical signal again and input into the initial signal input coding array. After the operation process simulates the optical path and completes k iterative operations, the kth iterative operation result read by the result reading optical path is used as the second largest eigenvalue and the corresponding second eigenvector.

13. A target data identification method, characterized in that: include: When receiving a data identification task of identifying target data from task data, obtaining task data related to the data identification task; The task data is expressed in a matrix form to obtain a matrix to be calculated, wherein the characteristic data of the matrix to be calculated is the target data; When it is determined that the characteristic data calculation process of the matrix to be calculated includes multiple iterative matrix-vector multiplication operations or multiple iterative matrix multiplication operations, the photon processor as described in any one of claims 1 to 12 is called to perform data processing on the matrix to be calculated to obtain the characteristic data of the matrix to be calculated as the processing result of the data recognition task.

14. The target data identification method according to claim 13, characterized in that: The data identification task is a hot event identification task, and the processing result of the data identification task is the data corresponding to the main eigenvalue of the matrix to be calculated.

15. The target data identification method according to claim 14, characterized in that: The task data is represented in a matrix form, including: Obtain each entity node, each preset time step and the corresponding preset time interval of the dynamic social network, and assign a unique index value to each entity node; Construct an initial matrix of matrix element values ​​to be filled according to the total number of entity nodes, wherein the row dimension and column dimension of the initial matrix are the same as the total number of entity nodes; In each preset time step, in each preset time interval of the current preset time step, the corresponding matrix element values ​​of the initial matrix are determined according to whether different entity nodes have a relationship, and the initial matrix filled with the matrix element values ​​is used as the matrix to be calculated.

16. The target data identification method according to claim 13, characterized in that: The data identification task is a data importance ranking task, and the process of generating the processing result of the data identification task includes: Obtaining each eigenvalue of the matrix to be calculated, and calculating the absolute value of each eigenvalue; Sort the eigenvalues ​​in descending or ascending order of absolute value to obtain feature sorting information; the absolute value of each eigenvalue is proportional to the importance of the corresponding data; The data corresponding to each eigenvalue is sorted according to the feature sorting information to obtain a sorting result corresponding to the data importance sorting task.

17. A target data identification device, characterized in that: include: A task data acquisition module, configured to, upon receiving a data recognition task of identifying target data from task data, acquire task data related to the data recognition task; A data processing module is used to express the task data in a matrix form to obtain a matrix to be calculated, wherein the characteristic data of the matrix to be calculated is the target data; A task execution module is used to, when it is determined that the characteristic data calculation process of the matrix to be calculated includes multiple iterative matrix-vector multiplication operations or multiple iterative matrix multiplication operations, perform data processing on the matrix to be calculated by calling the photon processor according to any one of claims 1 to 12, to obtain the characteristic data of the matrix to be calculated as the processing result of the data recognition task.

18. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the target data identification method according to any one of claims 13 to 16 when executing the computer program.

19. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a computer program, which, when executed by a processor, implements the steps of the target data identification method according to any one of claims 13 to 16.

20. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the target data identification method according to any one of claims 13 to 16 are implemented.

Images