Phase configuration method and electronic device

Abstract

The present disclosure provides a phase configuration method and electronic equipment, which relates to the technical field of communication, obtains a discrete point set corresponding to a phase interval of a first optical MZI array system, calibrates a first phase set obtained according to the discrete point set by using a test optical signal corresponding to the first optical MZI array system, adjusts each phase in the first optical MZI array system by using a second phase set, obtains a second optical MZI array system, and the second phase set includes at least one phase to be adjusted in the calibrated first phase set; obtains an optimal phase configuration corresponding to the second optical MZI array system according to a first output light intensity of each port in the second optical MZI array system under a continuous phase condition and a heuristic algorithm, and configures the phase of the second optical MZI array system by using the optimal phase configuration. The pre-calibration operation can be reduced, the power consumption can be reduced, the expressiveness and robustness of the system can be improved, the accuracy of the phase configuration can be improved, and resources can be saved.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a phase configuration method and an electronic device. Background Technology

[0002] With the development of science and technology and the increasing demand for computing power, new types of computing have emerged. Among them, optical computing, as a new computing paradigm, has developed rapidly and has a wide range of applications. For example, optical computing can perform optical calculations based on Mach-Zehnder interferometer (MZI) arrays, utilizing the interference principle of light waves to achieve efficient matrix operations. For instance, discretized phase modulation can alleviate noise problems in MZI array optical computing systems; however, discretized phase modulation, by locking a finite number of phase states, results in relatively poor accuracy in phase adjustment. Summary of the Invention

[0003] This disclosure provides a phase configuration method and an electronic device. Its main purpose is to solve the problem that discrete phase modulation, which locks a limited number of phase states, results in poor phase adjustment accuracy and requires additional devices.

[0004] According to a first aspect of this disclosure, a phase configuration method is provided, comprising:

[0005] Based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, obtain the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system, and obtain the first phase set based on the set of discrete points;

[0006] The first phase set is calibrated using the test optical signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain the calibrated first phase set.

[0007] Each phase in the second phase set of the first optical Mach-Zehnder interferometer array system is adjusted to obtain the second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted from the calibrated first phase set;

[0008] Obtain the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions;

[0009] Using a heuristic algorithm and a loss function corresponding to the first output light intensity, the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system is obtained, and the optimal phase configuration is used to perform phase configuration on the second optical Mach-Zehnder interferometer array system.

[0010] According to a second aspect of this disclosure, a phase configuration device is provided, comprising:

[0011] The set acquisition unit is used to acquire the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system according to the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, and acquire the first phase set according to the set of discrete points.

[0012] The system calibration unit is used to calibrate the first phase set using the test light signal corresponding to the first optical Mach-Zehnder interferometer array system, and obtain the calibrated first phase set.

[0013] A phase adjustment unit is used to adjust each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain a second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted in the calibrated first phase set;

[0014] The light intensity acquisition unit is used to acquire the first output light intensity of each port in the third optical Mach-Zehnder interferometer array system under continuous phase conditions.

[0015] The phase configuration unit is used to obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system by using a heuristic algorithm and a loss function corresponding to the first output light intensity, and to perform phase configuration on the second optical Mach-Zehnder interferometer array system using the optimal phase configuration.

[0016] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0017] At least one processor; and

[0018] A memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0020] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0021] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0022] This disclosure involves obtaining a set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, and obtaining a first phase set based on the set of discrete points; calibrating the first phase set using the test light signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain a calibrated first phase set; adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain a second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted from the calibrated first phase set; obtaining the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions; obtaining the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system using a heuristic algorithm and a loss function corresponding to the first output light intensity; and performing phase configuration on the second optical Mach-Zehnder interferometer array system using the optimal phase configuration. Therefore, phase can be discretized and calibrated. Continuous phase can be discretized into multiple pre-calibration endpoint values, requiring only the phase to be maintained near discrete points, reducing sensitivity to environmental disturbances. The discretization locking mechanism blocks phase drift propagation paths, avoiding noise accumulation and amplification issues in multi-stage cascaded calculations. Simultaneously, the pre-calibration mechanism replaces traditional real-time dynamic compensation, eliminating the hardware burden of high-frequency closed-loop feedback, significantly reducing system complexity and energy consumption, and achieving a "one-time calibration, long-term stability" operating mode. Phase configuration during pre-calibration eliminates the need for necessary real-time monitoring circuits and dynamic compensation modules in continuous phase systems at the hardware level, eliminating the need for additional calibration devices. This simplifies the number of optoelectronic components and control logic, reduces pre-calibration operations, lowers power consumption, improves system expressiveness and robustness, enhances phase configuration accuracy, and saves resources.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0024] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0025] Figure 1A A schematic diagram illustrating an example of a single MZI provided in an embodiment of this disclosure;

[0026] Figure 1B This is a schematic diagram illustrating an example of an MZI array provided in an embodiment of this disclosure;

[0027] Figure 2 This is a schematic diagram illustrating an example of a phase configuration method provided in an embodiment of this disclosure;

[0028] Figure 3 This is a schematic diagram illustrating another phase configuration method provided in an embodiment of the present disclosure;

[0029] Figure 4 This is a schematic diagram of an optical computation training method based on phase binary discretization error calibration provided in an embodiment of this disclosure;

[0030] Figure 5 This is a schematic diagram of an optical computing training system based on phase binary discretization error calibration provided in an embodiment of the present disclosure;

[0031] Figure 6 This is a schematic diagram of a phase configuration device provided in an embodiment of the present disclosure. Detailed Implementation

[0032] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0033] According to some embodiments, optical computing achieves a revolutionary breakthrough in physical principles through light as an information carrier. The massless and charge-free nature of light results in almost no energy loss during propagation in waveguides or free space, fundamentally breaking down the energy consumption wall of electronic interconnects. The multi-wavelength characteristics of light (wavelength division multiplexing) allow for the independent transmission of multiple signals in a single channel, enabling high-density parallel communication with bandwidths reaching the terahertz level, completely overcoming the speed and capacity limitations of electronic communication. The "propagation as computation" mechanism of optical computing can specifically utilize structures such as Mach-Zehnder interferometers and diffractive optical elements. During transmission, light directly performs matrix multiplication and convolution operations through physical effects such as interference and diffraction. This in-situ simulation computing mode integrates data processing into the optical path propagation, eliminating the need for data transport and naturally resolving the memory wall problem. Furthermore, the electromagnetic interference resistance and weak crosstalk of optical signals ensure the signal fidelity of large-scale parallel computing, while the ultra-low heat generation of optical computing alleviates heat dissipation pressure from the source.

[0034] Among them, optical computing based on Mach-Zehnder interferometer arrays utilizes the interference principle of light waves to achieve efficient matrix operations. Its core MZI unit consists of two optical couplers and two independently adjustable waveguide arms in the middle. By adjusting the refractive index (i.e., phase difference) of the waveguide arms through electrothermal or electro-optic effects, the coherent superposition state of the two beams can be precisely controlled, thereby achieving a linear transformation of light intensity at the output port. A large number of MZI units are interconnected in a grid-like topology, with each unit acting as a programmable matrix element; the entire array constitutes a reconfigurable optical matrix multiplier. When the input light signal propagates in parallel among multiple MZI stages, the natural interference process of the light waves synchronously completes the matrix and vector multiplication operations, eliminating the need for data transfer steps in traditional electronic devices. This in-situ computing method based on physical optics combines the advantages of ultra-high speed, low energy consumption, and high parallelism, making it particularly suitable for accelerating large-scale linear algebra operations and providing a new path to overcome the bottlenecks of traditional electronic computing.

[0035] Among them, such as Figure 1A As shown, a single MZI can be composed of two 50:50 beam splitters and two parallel waveguide arms, with one arm integrating two phase modulators (where a controllable phase difference is introduced). and ),like Figure 1A As shown, its transmission characteristics are described by a unitary matrix:

[0036] 1. First phase modulation matrix:

[0037] (1)

[0038] 2. First beam splitter transmission matrix:

[0039] (2)

[0040] 3. Second phase modulation matrix:

[0041] (3)

[0042] 4. Second beam splitter transmission matrix:

[0043] (4)

[0044] The unitary transmission matrix of a single MZI is:

[0045] (5)

[0046] By adjusting It can realize any 2×2 unitary transformation.

[0047] According to some embodiments, N MZI units can be cascaded in a specific topology (such as a triangular mesh) to form an MZI array, such as... Figure 1B As shown, for an N-dimensional input vector s, its light field representation E in After being injected into the N input ports, it undergoes a series of interference transformations within the array. For example... Figure 1B The dashed lines represent a single MZI (Multi-Unit Indicator). The transmission matrix of the entire network is the product of the unitary matrices of each MZI unit.

[0048] (6)

[0049] in It is important to note that this network can implement unitary transformations of any form because, according to Clems' decomposition theorem, any N×N unitary matrix V can be decomposed into:

[0050] (7)

[0051] Where D is the diagonal phase compensation matrix, derived from... Figure 1B The phase shifter is implemented outside the outermost dashed line. O represents a specific MZI ordering. All are precisely configured... This allows for programming to achieve the target unitary matrix V and output the light field E. out =VE in This is the result of matrix multiplication.

[0052] In MZI array-based optical computing, the input vector The data loading process is completed by injecting the optical signal into the input port through a spatial light modulator. The optical signal is in c / n eff velocity (n) eff The light (at the waveguide's effective refractive index) passes through the array, undergoes continuous interference transformations, and the transmission time is on the order of picoseconds. Finally, the light intensity is measured at the output end via a photodetector array to obtain the result.

[0053] According to some embodiments, slight changes in ambient temperature or mechanical stress can cause significant phase drift (e.g., a 0.1°C temperature change may cause a 10 milliradian phase change), thus introducing phase noise. Phase calibration, by monitoring output error in real time and dynamically adjusting to compensate for phase shift, can stabilize output accuracy in the short term. However, calibration requires additional optoelectronic components and closed-loop control, increasing system complexity and power consumption. Furthermore, it is difficult to eliminate the randomness of noise and the spatial correlation of environmental disturbances, resulting in inherent limitations in accuracy maintenance. Calibration for continuous signals requires frequent operations and continuous feedback. Maintaining accuracy requires high-cost monitoring plus dynamic compensation and real-time closed-loop feedback, increasing power consumption, and feedback delays can exacerbate system instability.

[0054] Phase discretization can alleviate the problems caused by noise to some extent, trading limited precision for a significant improvement in system stability. Discretized phase modulation locks a finite number of phase states, so the system only needs to keep the phase near discrete points. Small environmental disturbances will not trigger state transitions, and drift sensitivity is drastically reduced. At the same time, it does not require real-time calibration for all states, but only pre-calibration for a finite number of discrete states. It eliminates real-time feedback and only requires short-term periodic calibration, reducing hardware complexity and power consumption.

[0055] According to some embodiments, when training an optical neural network (ONN) using an MZI array, phase discretization improves stability but limits the precision of weight representation. Discretization compresses continuous phases into a finite number of states, causing the weights to be unable to accurately match the subtle adjustments required by gradient descent (e.g., 4 bits with only 16 discrete points cannot represent a change of 0.001 radians). Forced rounding to the nearest discrete point introduces accumulated quantization error, significantly reducing the model's expressive power and making it difficult for the model to converge to the optimal solution.

[0056] The phase configuration method and electronic device of the present disclosure are described below with reference to the accompanying drawings.

[0057] Figure 2 This is a schematic flowchart illustrating a phase configuration method provided in an embodiment of this disclosure. Figure 2 As shown, the method includes the following steps:

[0058] Step 101: Based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, obtain the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system, and obtain the first phase set based on the set of discrete points.

[0059] According to some embodiments, the implementing entity of this disclosure may be, for example, an electronic device. The name of the electronic device is not limited. For example, the electronic device may also be called a computer device, a terminal device, etc. The electronic device does not specifically refer to a particular fixed device. For example, when the structure of the electronic device changes, the electronic device may also change accordingly. For example, when the device identifier of the electronic device changes, the electronic device may also change accordingly.

[0060] According to some embodiments, the optical Mach-Zehnder interferometer array system can be, for example, an optical computing system including MZIs. This optical Mach-Zehnder interferometer array system is not specifically defined as a fixed system. For example, when the number of MZIs included in the optical Mach-Zehnder interferometer array system changes, the optical Mach-Zehnder interferometer array system can also change accordingly. For example, when a particular MZI in the optical Mach-Zehnder interferometer array system changes, the optical Mach-Zehnder interferometer array system can also change accordingly.

[0061] In some embodiments, the first optical Mach-Zehnder interferometer array system may refer, for example, to the system to be phase-configured. The term "first" in this first optical Mach-Zehnder interferometer array system is used to distinguish it from other optical Mach-Zehnder interferometer array systems and does not specifically refer to a particular fixed system. The first optical Mach-Zehnder interferometer array system may, for example, be a system that has already undergone coarse training.

[0062] In some embodiments, the requirement information may refer to the requirement information corresponding to the current optical MZI array system. This requirement information may include, for example, the recognition system of the current optical MZI array system or the recognition efficiency of the current optical MZI array system. This requirement information is not specifically defined by any fixed information. For example, when the array system of the optical MZI array system changes, the requirement information may also change accordingly.

[0063] In some embodiments, the phase interval may refer, for example, to the phase interval of an optical MZI array system. This phase interval may be, for example,... .

[0064] According to some embodiments, the set of discrete points can be, for example, a set of discrete points obtained by dividing the phase interval. This set of discrete points does not specifically refer to a fixed set. For example, when the way the phase interval is divided changes, the set of discrete points can also change accordingly.

[0065] In some of the following embodiments, for example, the midpoint of the interval corresponding to adjacent discrete points can be used as the phase value of the interval.

[0066] In some embodiments, based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, a set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system is obtained, and a first phase set is obtained based on the set of discrete points.

[0067] Step 102: The first phase set is calibrated using the test light signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain the calibrated first phase set;

[0068] According to some embodiments, the test optical signal may be, for example, an optical signal used to test an optical MZI array system. This optical signal is not specifically a fixed signal. For example, the test optical signal may change accordingly when its phase changes.

[0069] In some embodiments, the calibrated first phase set may be, for example, the set obtained after calibrating a first phase set. This calibrated phase set does not specifically refer to a fixed set. For example, the calibrated phase set may change accordingly when the calibration method changes. For instance, the calibrated phase set may also change accordingly when the test optical signal changes.

[0070] In some embodiments, the first phase set is calibrated using the test light signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain the calibrated first phase set.

[0071] Step 103: Adjust each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain the second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted in the calibrated first phase set;

[0072] According to some embodiments, the second phase set may, for example, include at least one phase to be adjusted from the calibrated first phase set. The "second" in the second phase set is used to distinguish it from the remaining phase sets and does not specifically refer to a fixed phase set. For example, the second phase set may also change accordingly when the phase to be adjusted changes. For example, the second phase set may also change accordingly when the number of phases to be adjusted corresponding to the second phase set changes.

[0073] In some embodiments, the second optical Mach-Zehnder interferometer array system may be, for example, a system obtained by adjusting the phases in the second phase set of the first optical Mach-Zehnder interferometer array system.

[0074] According to some embodiments, each phase in the second phase set of the first optical Mach-Zehnder interferometer array system is adjusted to obtain the second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted from the calibrated first phase set.

[0075] Step 104: Obtain the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions;

[0076] According to some embodiments, each port can be used, for example, to indicate a port in a second optical Mach-Zehnder interferometer array system. This port can be, for example, an output port. This output port can, for example, output light intensity. Each port does not specifically refer to a fixed port. For example, when the second optical Mach-Zehnder interferometer array system changes, the ports of the second optical Mach-Zehnder interferometer array system can also change accordingly.

[0077] According to some embodiments, a continuous phase condition may refer, for example, to the continuous, non-abrupt phase change during signal modulation. This continuous phase condition is not specifically defined by a fixed condition. For example, the continuous phase condition may change accordingly if the acquisition time point changes.

[0078] According to some embodiments, the first output light intensity may be, for example, the output light intensity corresponding to each port under the current continuous phase condition. This first output light intensity does not specifically refer to a fixed light intensity. For example, when the second optical Mach-Zehnder interferometer array system changes, the first output light intensity of each port may also change accordingly.

[0079] In some embodiments, the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions is obtained.

[0080] Step 105: Using a heuristic algorithm and the loss function corresponding to the first output light intensity, obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system, and use the optimal phase configuration to perform phase configuration on the second optical Mach-Zehnder interferometer array system.

[0081] According to some embodiments, a heuristic algorithm can be, for example, an algorithm based on empirical rules or heuristic information that quickly finds a feasible solution or a near-optimal solution by performing a heuristic search in the solution space of a problem. The heuristic algorithm in this disclosure can, for example, be an algorithm for obtaining the optimal phase configuration. This heuristic algorithm is not specifically defined as a fixed algorithm. For example, the heuristic algorithm can change accordingly when the algorithm type changes. For example, the heuristic algorithm can also change accordingly when the algorithm parameters change.

[0082] In some embodiments, a loss function is used to measure the difference between the model's predicted values ​​and the true values. This loss function can, for example, be used as the objective function of an optimization algorithm. The loss function is not specifically a fixed function. For example, the loss function can change when the way it is obtained changes. Similarly, the loss function can change when the parameters corresponding to it change.

[0083] According to some embodiments, the optimal phase configuration may be, for example, the optimal phase configuration corresponding to the current optical Mach-Zehnder interferometer array system. This optimal phase configuration is not specifically defined by a fixed configuration. For example, the optimal phase configuration may change when the method of obtaining it changes. Similarly, the optimal phase configuration may change when the heuristic algorithm changes.

[0084] In some embodiments, a heuristic algorithm and a loss function corresponding to the first output light intensity are used to obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system, and the optimal phase configuration is used to perform phase configuration on the second optical Mach-Zehnder interferometer array system.

[0085] This disclosure involves obtaining a set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, and obtaining a first phase set based on the set of discrete points; calibrating the first phase set using the test light signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain a calibrated first phase set; adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain a second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted in the calibrated first phase set; obtaining the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions; obtaining the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system using a heuristic algorithm and a loss function corresponding to the first output light intensity; and performing phase configuration on the second optical Mach-Zehnder interferometer array system using the optimal phase configuration. Therefore, phase can be discretized and calibrated. Continuous phase can be discretized into multiple pre-calibration endpoint values, requiring only the phase to be maintained near discrete points, reducing sensitivity to environmental disturbances. The discretization locking mechanism blocks phase drift propagation paths, avoiding noise accumulation and amplification issues in multi-stage cascaded calculations. Simultaneously, the pre-calibration mechanism replaces traditional real-time dynamic compensation, eliminating the hardware burden of high-frequency closed-loop feedback, significantly reducing system complexity and energy consumption, and achieving a "one-time calibration, long-term stability" operating mode. Phase configuration during pre-calibration eliminates the need for necessary real-time monitoring circuits and dynamic compensation modules in continuous phase systems at the hardware level, eliminating the need for additional calibration devices. This simplifies the number of optoelectronic components and control logic, reduces pre-calibration operations, lowers power consumption, improves system expressiveness and robustness, enhances phase configuration accuracy, and saves resources.

[0086] Furthermore, in one possible implementation of this embodiment, Figure 3This is a schematic flowchart illustrating another phase configuration method provided in an embodiment of this disclosure. Figure 3 As shown, the method includes the following steps:

[0087] Step 201: Based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, obtain the number of phase intervals corresponding to the first optical Mach-Zehnder interferometer array system.

[0088] The relevant processes can be described as above, and will not be repeated here.

[0089] According to some embodiments, Figure 4 This is a schematic diagram of an optical computation training method based on phase binary discretization error calibration provided in an embodiment of this disclosure. Figure 5 This is a schematic diagram of an optical computing training system based on phase binary discretization error calibration provided in an embodiment of this disclosure, as shown below. Figure 4 As shown, this method may include system initialization settings, dividing phase intervals and performing calibration, performing preliminary training, determining phase correlation values ​​and setting selectable phase values, testing output light intensity under discrete phases, calculating the loss function, solving for the optimal sequence, setting the final phase, and determining whether to end the training process based on the convergence result. Figure 5 As shown, the system can include an MZI array, a spatial light modulator, a photodetector array, external control equipment, and external computing equipment. The MZI array, as the core execution component, is composed of multiple MZI units cascaded in a triangular mesh or similar topological structure. Each unit contains a beam splitter and a waveguide arm with a phase shifter. The spatial light modulator is used to load the input vector onto the MZI array input. The photodetector array is used to measure the light intensity at the output port. The external control equipment is responsible for overall system control, such as setting the initial phase, dividing the phase interval, and running the algorithm. The external computing equipment assists in data processing, such as calculating the loss function. Among these devices, the optical MZI array system needs to include three operating modes: calibration mode, coarse training mode, and parameter fine-tuning mode. The calibration mode is used to calibrate discrete phase values ​​to eliminate errors; the coarse training mode is used for initial training using conventional methods; and the parameter fine-tuning mode is used for subsequent precise parameter adjustment.

[0090] The collaborative workflow between devices is as follows: After system startup, the external control device loads a test signal through the spatial light modulator to ensure the initial state of the phase shifter. The photodetector array measures the output light intensity and feeds it back to the external control device. Next, the external control device divides the phase interval, sets the MZI array into calibration mode, and inputs a test signal to complete calibration. Afterward, the external control device sets the system to coarse training mode and performs preliminary training through methods such as backpropagation until near convergence. Subsequently, it enters parameter fine-tuning mode, where the external control device determines the intermediate value and adjustment value of each phase interval, sets selectable phase values, loads input data through the spatial light modulator, transmits the optical signal through the MZI array, the photodetector array measures and records the output light intensity, the external computing device calculates the loss function between the ideal output and the discretized output, the external control device runs a simulated bifurcation algorithm to solve for the optimal sequence, and then sets the final phase based on this sequence. The above fine-tuning process is repeated until training is complete. Throughout the process, the devices work together to achieve the optical computing parameter design based on phase binary discretization through data transmission and control command exchange. A detailed description of each step can be provided below.

[0091] According to some embodiments, the method further includes:

[0092] The test light signal is input to the third optical Mach-Zehnder interferometer array system, and a preset vector is loaded through the spatial light modulator to control the state of the phase shifters of each Mach-Zehnder interferometer unit in the third optical Mach-Zehnder interferometer array system so that the phase shifters of each Mach-Zehnder interferometer unit are in the preset state.

[0093] According to some embodiments, the first optical Mach-Zehnder interferometer array system may be, for example, a system obtained after coarsely training a third optical Mach-Zehnder interferometer array system. This third optical Mach-Zehnder interferometer array system does not specifically refer to a fixed system. For example, when the third optical Mach-Zehnder interferometer array system changes, the first optical Mach-Zehnder interferometer array system may also change accordingly.

[0094] The preset state could be, for example, the phase of the phase shifter in each Mach-Zehnder interferometer unit being a preset value. This preset state is not specifically a fixed state. For example, when the preset phase value changes, the preset state can also change accordingly.

[0095] According to some embodiments, a spatial light modulator (SLM) can be, for example, a programmable device capable of actively controlling the modulation of a parameter of a light field, such as amplitude, phase, or polarization state, by liquid crystal molecules, thereby writing information into a light wave. This spatial light modulator does not specifically refer to a fixed modulator; for example, when the structure corresponding to the spatial light modulator changes, the spatial light modulator can also change accordingly.

[0096] Specifically, during the initial system setup phase, for example, the optical MZI array system can be started, and a test light signal can be input through an external control device. A simple vector, i.e., a preset vector, is loaded through a spatial light modulator to ensure that all phase shifters are in the preset state of the external control device. Here, the simple vector can be an all-one vector, and the preset state can be, for example, the phase. Set to known default values. The light intensity measured at the output port of the connected photodetector array can be returned to an external control device. This optical MZI array system can include calibration mode, coarse training mode, and parameter fine-tuning mode.

[0097] Step 202: Divide the phase interval of the first optical Mach-Zehnder interferometer array system according to the number of parts, and obtain the set of discrete points corresponding to the phase interval. The range between two adjacent discrete points is a phase sub-interval.

[0098] The relevant processes can be described as above, and will not be repeated here.

[0099] According to some embodiments, phase intervals can be divided and calibrated. Specifically, the 0~2π phase interval is divided into N equal parts on an external control device, resulting in discrete points of 0, 2π / N, 2×2π / N, ..., 2π. The range between two adjacent discrete points constitutes an interval. Where n... i Let N be a non-negative integer not greater than N. The midpoint of this interval can be denoted as . .in, It can be set to ,in, The middle value of the interval. This indicates the direction of the adjustment, i.e., whether to set it to the left or right endpoint value. This is the adjusted value.

[0100] The value of N can be determined based on the requirements. For example, a larger value can ensure the expressive power of the model and can be considered a small value. Compared to continuous phase settings, which require recalibration for each encountered phase, this discretized phase can be calibrated before training and then the calibration value can be directly applied, increasing the robustness of the system and improving the efficiency of operation.

[0101] According to some embodiments, the first in the MZI array Each MZI unit contains two phase shifters, such as Figure 1A As shown, the phase of the first phase shifter is The value after discretization is The phase of the second phase shifter is And the value after discretization is Therefore, the transmission matrix of the MZI can be represented as shown in formula (8):

[0102] (8)

[0103] Here, it is assumed that the input of the MZI is Therefore, its output under the first-order approximation should be:

[0104]

[0105] Note that the first item enclosed in parentheses in the above matrix is ​​the output of the MZI when the phase is set to the midpoint value of the interval. This can be... Let the coefficient matrix be denoted as b, and the output when the phase is set to the midpoint of the interval be denoted as B. Then, under the first-order approximation, the output is:

[0106] (9)

[0107] Formula (9) means that the output after discretization can be regarded as the output with the phase taken as the midpoint of the interval plus some small determinations to determine whether to take the upper or lower limit.

[0108] Step 203: The first phase set is calibrated using the test optical signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain the calibrated first phase set;

[0109] The relevant processes can be described as above, and will not be repeated here.

[0110] According to some embodiments, when the first optical Mach-Zehnder interferometer array system is set to calibration mode via an external control device, a test optical signal can be input to calibrate these N phase values ​​to eliminate phase shifter background errors and beam splitter errors. Therefore, these discrete points can be pre-calibrated to eliminate system errors. This mechanism utilizes pre-calibration to suppress phase drift caused by environmental disturbances, while replacing real-time dynamic compensation, significantly reducing hardware complexity and energy consumption.

[0111] In some embodiments, the phase sub-interval may be, for example, multiple sub-intervals obtained by dividing the phase interval.

[0112] According to some embodiments, the method further includes:

[0113] The third optical Mach-Zehnder interferometer array system was trained using backpropagation. The first optical Mach-Zehnder interferometer array system was obtained when the output data of the third optical Mach-Zehnder interferometer array system met the data requirements.

[0114] According to some embodiments, when using a third optical MZI array system to train a neural network, the third optical MZI array system can first be set to coarse training mode through an external control system. Backpropagation and other methods can be used to train the third optical MZI array system in the early stage until the error of the output result of the third optical MZI array system no longer decreases significantly, that is, it is close to the convergence state, and the first optical Mach-Zehnder interferometer array system can be obtained.

[0115] Step 204: Adjust each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain the second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted in the calibrated first phase set;

[0116] The relevant processes can be described as above, and will not be repeated here.

[0117] According to some embodiments, adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain the second optical Mach-Zehnder interferometer array system includes:

[0118] Obtain the phase sub-intervals corresponding to each phase in the second phase set of the first optical Mach-Zehnder interferometer array system;

[0119] Obtain the median value corresponding to the phase sub-interval, and obtain the adjustment value corresponding to the phase sub-interval;

[0120] The second optical Mach-Zehnder interferometer array system is obtained by adjusting each phase using intermediate and adjustment values. Therefore, adjusting the phase using adjustment and intermediate values ​​can improve the accuracy of phase acquisition, the accuracy of system adjustment, and the accuracy of system adjustment across applications.

[0121] According to some embodiments, each phase is adjusted using intermediate and adjustment values ​​to obtain a second optical Mach-Zehnder interferometer array system, including:

[0122] Obtain the adjustment direction information corresponding to each phase, where the adjustment direction information is a binary vector;

[0123] The second optical Mach-Zehnder interferometer array system is obtained by adjusting each phase using adjustment direction information, intermediate values, and adjustment values. Therefore, the phase adjustment direction can be determined by adjusting the direction information, thereby improving the accuracy of phase adjustment and the overall accuracy of the optical Mach-Zehnder interferometer array system adjustment.

[0124] According to some embodiments, the adjustment value corresponding to the phase sub-interval is obtained, including:

[0125] Based on the number of segments and the phase interval, obtain the adjustment value corresponding to the phase sub-interval.

[0126] In some embodiments, each phase that needs to be adjusted in the second optical Mach-Zehnder interferometer array system can be determined on an external control device. (where phase) It can represent and ), determine the interval in which the phase lies, if it falls within the nth interval. i +1 intervals (counting from 0), calculate the median value of each interval. Simultaneously calculate the adjustment value .

[0127] In some embodiments, the optical MZI array system can be set to a parameter fine-tuning mode via an external control device, allowing adjustment of each phase. Set as , using symbols Indicates the direction of selection, when Time selection ,when Time selection Among them, can be The phase value was set to the calibrated value.

[0128] Step 205: Obtain the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions;

[0129] The relevant processes are as described above, and will not be repeated here.

[0130] According to some embodiments, the output light intensity under discrete phase can be obtained. When the phase in the MZI unit is loaded into the input terminal of the MZI array through a spatial light modulator by an external control device, the light signal is allowed to pass through the array, and then the light intensity of each output port is measured and recorded by a photodetector, the second output light intensity can be obtained.

[0131] According to some embodiments, the method further includes:

[0132] Obtain the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under discrete phase conditions;

[0133] Based on the first and second output light intensities, the loss function corresponding to the first output light intensity is obtained. Therefore, the loss function can be obtained through the output light intensities of each port, improving the accuracy of the loss function acquisition.

[0134] According to some embodiments, a loss function can be calculated. For example, the first output light intensity under ideal conditions can be calculated on an external computing device under continuous phase conditions, and at the same time, the difference between the output light intensity under ideal conditions and the second output light intensity under discretized phase conditions can be calculated on the external computing device to obtain the loss function.

[0135] According to some embodiments, a loss function corresponding to the first output light intensity is obtained based on the first output light intensity and the second output light intensity, including:

[0136] Obtain the difference between the first and second output light intensities of each port;

[0137] The difference values ​​at each port in the second optical Mach-Zehnder interferometer array system are summed to obtain the loss function corresponding to the first output light intensity.

[0138] According to some embodiments, it is assumed that the first The input of each MZI is As can be seen from the above analysis, This represents the offset of the beam of light relative to the midpoint of the interval. For example, it could be the input described above... It is written as the output of the midpoint value of the phase interval plus a certain offset, which comes from the result of phase discretization processing of the passed MZI cells. At this point, it can be written as a linear combination of the offsets of the passed MZI cells. So after the first After one MZI unit, a new type of MZI unit will be introduced. New deviations, if the MZI array has A phase shifter and If there are one output port, then the overall output of the MZI array is:

[0139]

[0140] in, This refers to the output when the phase is set to the midpoint of the interval. Similarly, without phase discretization, the ideal output would be:

[0141]

[0142] in, For the first The difference between each phase and the midpoint of its interval. This ideal output can be expressed by the formula... The ideal MZI array matrix change is calculated. Similarly, since the training is close to convergence at this point, calculating the output light intensity under ideal conditions will not take up too much system time.

[0143] After obtaining the output after phase discretization and the output without discretization, the difference between them can be directly compared. Specifically, in optical computing, the measurement formula... and The complex amplitude in the MZI array is quite complex and can be used for light intensity measurement. Specifically, consider the first... One output, And the ideal case of the first Output of each port Its output light intensity, in the lowest order approximation, is...

[0144]

[0145] Among them, can Recorded as Similarly, for the output light intensity under ideal conditions, the lowest-order approximation should be:

[0146]

[0147] By comparing the differences in output light intensity between the two and summing the results over all ports, we can obtain:

[0148]

[0149] Ignore minimization After removing the ineffective constant terms, the loss function can be obtained. ,

[0150]

[0151] If let ,and Therefore, the loss function can also be rewritten as:

[0152]

[0153] at this time, It is a binary variable, taking values ​​of At this time This is a standard Ising model. A set of... The value of makes the loss function Minimum. The task of training an optical neural network can be transformed into finding the ground state of the Ising model. This process can be carried out using heuristic algorithms such as simulated bifurcation algorithms. When using heuristic algorithms, gradient information does not need to be calculated, thus avoiding the problems caused by gradient loss in backpropagation. Therefore, by setting the phase as the endpoint value of the interval and transforming the difference between the output light intensity and the ideal value into the Ising model, a global search for the optimal discrete phase combination is performed using a heuristic algorithm, making the discretized output approximate the accuracy of continuous parameters.

[0154] Wherein, coupling coefficient It can be derived from the output light intensity error, specifically as follows: ,in This is the correlation coefficient representing the real part of the complex amplitude of the output optical field. (External field coefficient) It can be derived from the first-order term of the output light intensity error, specifically as follows: ,in This represents the difference between the phase and the midpoint of the interval under ideal output conditions.

[0155] Step 206: Using a heuristic algorithm and the loss function corresponding to the first output light intensity, obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system, and use the optimal phase configuration to perform phase configuration on the second optical Mach-Zehnder interferometer array system.

[0156] The relevant processes can be described as above, and will not be repeated here.

[0157] According to some embodiments, the optimal phase configuration may be, for example, optimal The sequence. The loss function obtained above can be processed using an external control device based on a heuristic algorithm. Initially, all settings are... For rounding phase The value is then calculated iteratively according to the steps suggested by the heuristic algorithm, and the result is updated. The value, until all All fixed in or Record this The sequence is the optimal one. sequence.

[0158] According to some embodiments, optimal methods can be adopted. The sequence sets the final phase for the two phase modulators of each MZI unit, i.e. .

[0159] According to some embodiments, the above steps can be repeated until training is complete.

[0160] According to some embodiments, in the simulated bifurcation algorithm, it is assumed that the current state is in the update phase. Thus making The smallest Step, then in During the step, each The value of should be,

[0161]

[0162]

[0163] in, For time interval variables, For time evolution variables, is the time evolution coefficient, a positive number. When arrive After that, it remained at Keep the top still and move the corresponding Set to This solution method is a heuristic approach, utilizing the fact that... At that time, the stable point of the system is such that the formula Find the point with the smallest value. Then, when each... After evolution stops, by keeping it in Binarize it to obtain a set of values. of Sequence. When obtained Then, through the formula , This will give you all the phase settings.

[0164] According to some embodiments, a heuristic algorithm is used to process the loss function corresponding to the first output light intensity to obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system, including:

[0165] Based on each phase in the second phase set, obtain the first binary variable sequence;

[0166] The first binary variable sequence is iteratively processed using a heuristic algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, the second binary variable sequence is obtained, and the second binary variable sequence is used as the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system.

[0167] According to some embodiments, a heuristic algorithm and a loss function corresponding to the first output light intensity are used to iteratively process the first binary variable sequence. When all variables in the first binary variable sequence converge to a preset value, a second binary variable sequence is obtained, including:

[0168] The first binary variable sequence is iteratively processed using a simulated bifurcation algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, the second binary variable sequence is obtained.

[0169] According to some embodiments, the phase configuration of the second optical Mach-Zehnder interferometer array system is performed using optimal phase configuration, including:

[0170] Based on the optimal phase configuration, the median value of each phase sub-interval, and the adjustment value, the two phase modulators of each Mach-Zehnder interferometer unit in the second optical Mach-Zehnder interferometer array system are configured in phase.

[0171] According to some embodiments, the method further includes:

[0172] If the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system is not obtained when using a heuristic algorithm and the loss function corresponding to the first output light intensity, the step of adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system is re-executed to obtain the second optical Mach-Zehnder interferometer array system.

[0173] In one or related embodiments, based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, the number of segments corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system is obtained; based on the number of segments, the phase interval of the first optical Mach-Zehnder interferometer array system is divided to obtain the set of discrete points corresponding to the phase interval, wherein the range between two adjacent discrete points is a phase sub-interval. The number of segments can be determined according to the requirement information to improve the accuracy of phase interval division and the accuracy of phase configuration.

[0174] According to embodiments of this disclosure, a phase configuration device is also provided.

[0175] For example, Figure 6 This is a schematic diagram of a phase configuration device provided in an embodiment of the present disclosure. The phase configuration device 600 includes: a collection acquisition unit 601, a system calibration unit 602, a phase adjustment unit 603, a light intensity acquisition unit 604, and a phase configuration unit 605; wherein,

[0176] The set acquisition unit 601 is used to acquire the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system according to the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, and acquire the first phase set according to the set of discrete points.

[0177] The system calibration unit 602 is used to calibrate the first phase set using the test optical signal corresponding to the first optical Mach-Zehnder interferometer array system, and to obtain the calibrated first phase set.

[0178] Phase adjustment unit 603 is used to adjust each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain the second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted in the calibrated first phase set;

[0179] The light intensity acquisition unit 604 is used to acquire the first output light intensity of each port in the third optical Mach-Zehnder interferometer array system under continuous phase conditions.

[0180] The phase configuration unit 605 is used to obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system by using a heuristic algorithm and a loss function corresponding to the first output light intensity, and to perform phase configuration on the second optical Mach-Zehnder interferometer array system using the optimal phase configuration.

[0181] Furthermore, the set acquisition unit 601, when acquiring the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, is specifically used for:

[0182] Based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, obtain the number of phase intervals corresponding to the first optical Mach-Zehnder interferometer array system.

[0183] Based on the number of parts, the phase interval of the first optical Mach-Zehnder interferometer array system is divided to obtain the set of discrete points corresponding to the phase interval. The range between two adjacent discrete points is a phase sub-interval.

[0184] Furthermore, the phase adjustment unit 603 is used to adjust each phase in the second phase set of the first optical Mach-Zehnder interferometer array system. Specifically, when acquiring the second optical Mach-Zehnder interferometer array system, it is used for:

[0185] Obtain the phase sub-intervals corresponding to each phase in the second phase set of the first optical Mach-Zehnder interferometer array system;

[0186] Obtain the median value corresponding to the phase sub-interval, and obtain the adjustment value corresponding to the phase sub-interval;

[0187] The second optical Mach-Zehnder interferometer array system was obtained by adjusting each phase using intermediate and adjustment values.

[0188] Furthermore, the phase adjustment unit 603 is used to adjust each phase using intermediate and adjustment values. Specifically, when acquiring the second optical Mach-Zehnder interferometer array system, it is used for:

[0189] Obtain the adjustment direction information corresponding to each phase, where the adjustment direction information is a binary vector;

[0190] The second optical Mach-Zehnder interferometer array system is obtained by adjusting the direction information, intermediate values, and adjustment values ​​to each phase.

[0191] Furthermore, the phase adjustment unit 603, when acquiring the adjustment value corresponding to the phase sub-interval, is specifically used for:

[0192] Based on the number of segments and the phase interval, obtain the adjustment value corresponding to the phase sub-interval.

[0193] Furthermore, the phase configuration unit 605 is specifically used for:

[0194] Obtain the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under discrete phase conditions;

[0195] Based on the first output light intensity and the second output light intensity, obtain the loss function corresponding to the first output light intensity.

[0196] Furthermore, the phase configuration unit 605, when acquiring the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under discrete phase conditions, is specifically used for:

[0197] The adjusted second phase set is loaded onto the input end of the optical Mach-Zehnder interferometer array in the second optical Mach-Zehnder interferometer array system, and the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under discrete phase conditions is measured using a photodetector.

[0198] Furthermore, the phase configuration unit 605, when obtaining the loss function corresponding to the first output light intensity based on the first output light intensity and the second output light intensity, is specifically used for:

[0199] Obtain the difference between the first and second output light intensities of each port;

[0200] The difference values ​​at each port in the second optical Mach-Zehnder interferometer array system are summed to obtain the loss function corresponding to the first output light intensity.

[0201] Furthermore, the phase configuration unit 605, when processing the loss function corresponding to the first output light intensity using a heuristic algorithm to obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system, is specifically used for:

[0202] Based on each phase in the second phase set, obtain the first binary variable sequence;

[0203] The first binary variable sequence is iteratively processed using a heuristic algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, the second binary variable sequence is obtained, and the second binary variable sequence is used as the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system.

[0204] Furthermore, the phase configuration unit 605 is used to iteratively process the first binary variable sequence using a heuristic algorithm and a loss function corresponding to the first output light intensity. When the second binary variable sequence is obtained after all variables in the first binary variable sequence converge to a preset value, it is specifically used for:

[0205] The first binary variable sequence is iteratively processed using a simulated bifurcation algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, the second binary variable sequence is obtained.

[0206] Furthermore, the phase configuration unit 605, when performing phase configuration of the second optical Mach-Zehnder interferometer array system using optimal phase configuration, is specifically used for:

[0207] Based on the optimal phase configuration, the median value of each phase sub-interval, and the adjustment value, the two phase modulators of each Mach-Zehnder interferometer unit in the second optical Mach-Zehnder interferometer array system are configured in phase.

[0208] Furthermore, the phase configuration unit 605 is specifically used for:

[0209] If the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system is not obtained when using a heuristic algorithm and the loss function corresponding to the first output light intensity, the step of adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system is re-executed to obtain the second optical Mach-Zehnder interferometer array system.

[0210] Furthermore, the phase configuration unit 605 is specifically used for:

[0211] The test light signal is input to the third optical Mach-Zehnder interferometer array system, and a preset vector is loaded through the spatial light modulator to control the state of the phase shifters of each Mach-Zehnder interferometer unit in the third optical Mach-Zehnder interferometer array system so that the phase shifters of each Mach-Zehnder interferometer unit are in the preset state.

[0212] Furthermore, the phase configuration unit 605 is specifically used for:

[0213] The third optical Mach-Zehnder interferometer array system was trained using backpropagation. The first optical Mach-Zehnder interferometer array system was obtained when the output data of the third optical Mach-Zehnder interferometer array system met the data requirements.

[0214] It should be noted that the description of the features in the embodiment corresponding to the phase configuration device can be found in the relevant description of the embodiment corresponding to the phase configuration method, and will not be repeated here.

[0215] Embodiments of this disclosure also provide an electronic device including a memory and a processor, the memory storing a computer program, the processor being configured to run the computer program to perform the steps in any of the phase configuration method embodiments described above.

[0216] Embodiments of this disclosure also provide a computer-readable storage medium storing a computer program configured to execute the steps in any of the phase configuration method embodiments described above when the computer program is run.

[0217] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0218] Embodiments of this disclosure also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the phase configuration method embodiments described above.

[0219] Embodiments of this disclosure also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in any of the phase configuration method embodiments described above.

[0220] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0221] The phase configuration method provided by this disclosure has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this disclosure. The descriptions of the embodiments above are only for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make several improvements and modifications to this disclosure without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this disclosure.

Claims

1. A method of phase configuration of an optical computing system, characterized in that, include: Based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, obtain the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system, and obtain the first phase set based on the set of discrete points; The first phase set is calibrated using the test optical signal corresponding to the first optical Mach-Zehnder interferometer array system to obtain the calibrated first phase set. Each phase in the second phase set of the first optical Mach-Zehnder interferometer array system is adjusted to obtain the second optical Mach-Zehnder interferometer array system, wherein the second phase set includes at least one phase to be adjusted from the calibrated first phase set; Obtain the first output light intensity of each port in the second optical Mach-Zehnder interferometer array system under continuous phase conditions; Using a heuristic algorithm and a loss function corresponding to the first output light intensity, the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system is obtained, and the optimal phase configuration is used to perform phase configuration on the second optical Mach-Zehnder interferometer array system; The step of adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain the second optical Mach-Zehnder interferometer array system includes: Obtain the phase sub-intervals corresponding to each phase in the second phase set of the first optical Mach-Zehnder interferometer array system; Obtain the intermediate value corresponding to the phase sub-interval, and obtain the adjustment value corresponding to the phase sub-interval; The intermediate value and the adjustment value are used to adjust each phase to obtain the second optical Mach-Zehnder interferometer array system; The step of adjusting each phase using the intermediate value and the adjustment value to obtain the second optical Mach-Zehnder interferometer array system includes: Obtain the adjustment direction information corresponding to each phase, wherein the adjustment direction information is a binary vector; The adjustment direction information, the intermediate value, and the adjustment value are used to adjust each phase to obtain a second optical Mach-Zehnder interferometer array system; The method further includes: Obtain the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under discrete phase conditions; Based on the first output light intensity and the second output light intensity, obtain the loss function corresponding to the first output light intensity.

2. The method of claim 1, wherein, The step of obtaining the set of discrete points corresponding to the phase interval of the first optical Mach-Zehnder interferometer array system based on the requirement information of the first optical Mach-Zehnder interferometer array system includes: Based on the requirement information corresponding to the first optical Mach-Zehnder interferometer array system, obtain the number of phase intervals corresponding to the first optical Mach-Zehnder interferometer array system. Based on the number of parts, the phase interval of the first optical Mach-Zehnder interferometer array system is divided to obtain the set of discrete points corresponding to the phase interval, wherein the range between two adjacent discrete points is a phase sub-interval.

3. The method of claim 1, wherein, The step of obtaining the adjustment value corresponding to the phase sub-interval includes: Based on the number of segments corresponding to the phase interval and the phase interval, the adjustment value corresponding to the phase sub-interval is obtained.

4. The method of claim 1, wherein, The acquisition of the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under the discrete phase condition includes: The adjusted second phase set is loaded onto the input end of the optical Mach-Zehnder interferometer array in the second optical Mach-Zehnder interferometer array system, and the second output light intensity of each port in the second optical Mach-Zehnder interferometer array system under discrete phase conditions is measured using a photodetector.

5. The method of claim 1, wherein, The step of obtaining the loss function corresponding to the first output light intensity based on the first output light intensity and the second output light intensity includes: Obtain the difference between the first output light intensity and the second output light intensity of each port; The difference values ​​corresponding to each port in the second optical Mach-Zehnder interferometer array system are summed to obtain the loss function corresponding to the first output light intensity.

6. The method of claim 1, wherein, The step of processing the loss function corresponding to the first output light intensity using a heuristic algorithm to obtain the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system includes: Based on each phase in the second phase set, obtain the first binary variable sequence; The first binary variable sequence is iteratively processed using a heuristic algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, a second binary variable sequence is obtained, and the second binary variable sequence is used as the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system.

7. The method of claim 6, wherein, The first binary variable sequence is iteratively processed using a heuristic algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, a second binary variable sequence is obtained, including: The first binary variable sequence is iteratively processed using a simulated bifurcation algorithm and a loss function corresponding to the first output light intensity. When all variables in the first binary variable sequence converge to a preset value, the second binary variable sequence is obtained.

8. The method of claim 1, wherein, The step of performing phase configuration on the second optical Mach-Zehnder interferometer array system using the optimal phase configuration includes: Based on the optimal phase configuration, the median value of each phase sub-interval, and the adjustment value, the two phase modulators of each Mach-Zehnder interferometer unit in the second optical Mach-Zehnder interferometer array system are configured in phase.

9. The method of claim 1, wherein, The method further includes: If the optimal phase configuration corresponding to the second optical Mach-Zehnder interferometer array system is not obtained by using the heuristic algorithm and the loss function corresponding to the first output light intensity, the step of adjusting each phase in the second phase set of the first optical Mach-Zehnder interferometer array system to obtain the second optical Mach-Zehnder interferometer array system is re-executed.

10. The method of claim 1, wherein, The method further includes: The input test optical signal is input to a third optical Mach-Zehnder interferometer array system, and a preset vector is loaded through a spatial light modulator to control the state of the phase shifter of each Mach-Zehnder interferometer unit in the third optical Mach-Zehnder interferometer array system, so that the phase shifter of each Mach-Zehnder interferometer unit is in a preset state.

11. The method of claim 10, wherein, The method further comprises; The third optical Mach-Zehnder interferometer array system is trained in a back propagation manner, and the first optical Mach-Zehnder interferometer array system is obtained when the output data of the third optical Mach-Zehnder interferometer array system meets the data requirement.

12. An electronic device, comprising: Comprise: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the phase configuration method of any one of claims 1 to 11.

Images