A multi-dimensional expandable matrix operation system and its control method

Abstract

The present invention relates to the technical field of optical computing, and more specifically, to a multi-dimensional expandable matrix operation system and a control method thereof. The present invention loads the input matrix on light waves of different wavelengths, utilizes the wavelength dimension of light to expand the scale number of matrix operations, inputs the wavelength-division multiplexed optical intensity signals to different microring modulators, utilizes space-division multiplexing to show the expandability of matrix operation units, converts the matrix information into a time series, through time-division multiplexing, improves the bandwidth utilization rate and also expands the matrix operation scale, and at the same time supports real-time training and flexibly regulates the information of each matrix element. The present invention can flexibly control the input of matrix operations by using each dimension of light, can process data in parallel and accelerate, and can quickly solve large-scale problems.

Description

Technical Field

[0001] The present invention relates to the technical field of optical computing, and more particularly, to a multi-dimensional scalable matrix operation system and a control method thereof. Background Art

[0002] In recent years, with the continuous rise of artificial intelligence technology, the amount of data generated has increased exponentially, which has overloaded the transmission network, caused latency problems and increased power consumption. Therefore, there is a need to be able to achieve flexible multi-mode data processing, and at the same time have high computing throughput and energy efficiency. Integrated photonics provides unique potential advantages, including low interconnect loss and huge bandwidth, and is a promising platform for developing optical computing accelerators. Photonic devices can utilize different characteristics of light waves in time, wavelength and space, thereby improving parallelism and data throughput. The progress of photonic linear processors has further improved the data reuse efficiency and significantly reduced the energy overhead associated with optoelectronic and analog-to-digital conversion.

[0003] In actual operation, how to effectively utilize these optical wave sizes to expand the scale of the optical matrix processor has become a widely concerned solution. The research on optical matrix operation based solely on spatial dimension multiplexing realizes scalability by using space-division multiplexing (SDM), but it is difficult to solve large-scale problems because it requires a quadratic growth of components, which complicates manufacturing. In contrast, single wavelength-division multiplexing (WDM) reduces component requirements but relies on large dispersion elements. Single time-division multiplexing (TDM) serializes data into time waveforms, improves bandwidth utilization and supports real-time training, but cannot achieve low latency. However, the computational scale expansion of these architectures usually depends on spatial combinations with different time delays, so additional on-chip delay lines and dispersion elements are also required, which limits the compactness and flexibility of the photonic processor. Summary of the Invention

[0004] An object of the present invention is to overcome the deficiencies that the single-dimensional multiplexing of the prior art cannot achieve parallel large-scale matrix operations and at the same time limits the scale and number of devices, and to provide a multi-dimensional scalable matrix operation system and a control method thereof, which can perform large-scale matrix operations in parallel and achieve the scalability of matrix operations.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is:

[0006] Provide a multi-dimensional scalable matrix operation system, including:

[0007] A tunable laser: used to generate optical carrier signals with multiple carrier wavelengths;

[0008] Intensity modulator: used to modulate an optical carrier signal according to a first linear sequence in the time domain, load the optical carrier signal on the optical intensity, and obtain an optical intensity signal carrying intensity information;

[0009] Wavelength division multiplexer: used to multiplex optical intensity signals of different wavelengths and input them into the microring modulator;

[0010] Microring modulator: used to modulate the optical intensity signal according to a second linear sequence in the time domain to obtain a modulated signal; and superimpose the modulated signals of different symbol periods through microring feedback to obtain an output optical signal; the output optical signal carries the matrix multiplication and addition operation result of the first linear sequence and the second linear sequence;

[0011] Demultiplexer: used to separate output optical signals of different wavelengths onto each channel;

[0012] Photodetector: used to detect the intensity of the output optical signal on each channel.

[0013] A multi-dimensional expandable matrix operation system provided by the present invention loads the input matrix on optical waves of different wavelengths, utilizes the wavelength dimension of light to expand the scale number of matrix operations. Inputting the wavelength-division multiplexed optical intensity signals onto different microring modulators utilizes spatial division multiplexing to show the expandability of matrix operation units. Converting the matrix information into a time sequence, through time division multiplexing, improves the bandwidth utilization rate and also expands the matrix operation scale. At the same time, it supports real-time training and flexibly regulates the information of each matrix element. The present invention can flexibly control the input of matrix operations using each dimension of light, can process data in parallel and accelerate, and quickly solve large-scale problems.

[0014] Further, the first linear sequence is a time sequence formed by arranging the input matrix in the corresponding multiplication order; multiple input matrices form multiple parallel time sequences, that is, multiple first linear sequences, in the same order at the same time; the multiple first linear sequences are encoded onto different optical carrier wavelengths by the multiple intensity modulators; the second linear sequence is a time sequence formed by arranging the multiplication matrix in the multiplication order corresponding to the input matrix.

[0015] Further, the input matrix is arranged in row order to form the first linear sequence; the multiplication matrix is arranged in column order to form the second linear sequence.

[0016] Further, the microring modulator includes a 2×2 first multimode interferometer, a second multimode interferometer, and a Mach-Zehnder electro-optic modulator. The first multimode interferometer, the Mach-Zehnder electro-optic modulator, and the second multimode interferometer are cascaded, and the first input port of the first multimode interferometer is connected to the first output port of the second multimode interferometer to form a microring. The second input port of the first multimode interferometer is connected to the output end of the wavelength division multiplexer, and the second output port of the second multimode interferometer is connected to the input port of the photodetector. The arm lengths of the Mach-Zehnder interferometer are controlled by regulating electrodes for intensity modulation.

[0017] Further, the microring modulator is integrated on a thin-film lithium niobate substrate. In the present invention, the scalable matrix operation system can be integrated on a chip, enabling integration to expand computational parallelism or throughput, thereby obtaining a higher data volume.

[0018] Further, the intensity modulator is a Mach-Zehnder electro-optic modulator.

[0019] The present invention also provides a method for regulating a multi-dimensional scalable matrix operation system, including the following steps:

[0020] S1. Using a tunable laser to output optical carrier signals of multiple different wavelengths;

[0021] S2. Arranging the input matrix in a time series in the time domain according to the corresponding multiplication sequence to obtain a first linear sequence, and using an intensity modulator to encode the first linear sequence on optical signals of different wavelengths to obtain optical intensity signals carrying intensity information;

[0022] S3. Loading the optical intensity signals on different wavelength channels onto the same channel through a wavelength division multiplexer and inputting them into different microring modulators;

[0023] S4. Arranging the multiplication matrix in a time series in the time domain according to the multiplication order corresponding to the input matrix to obtain a second linear sequence; modulating the microring modulator according to the second linear sequence and in the time division multiplexing manner to obtain a modulated signal; using the delay of the microring to control the superposition of the modulated signals of different symbol periods to obtain an output optical signal, and controlling the output optical signal to output to the wavelength division multiplexer;

[0024] S5. Separating the output optical signals of different wavelengths onto each channel through a demultiplexer and controlling the output of the photodetector.

[0025] In the present invention, multiple wavelengths output by a tunable laser are made the same as the resonance wavelengths of the microrings. Through precise dispersion management of the Mach-Zehnder electro-optic modulator and the ring waveguide in the microring modulator, multiple carrier wavelengths are aligned with their optical resonances to maintain the same modulation response across the entire wavelength range. Controlling the tunable laser to output multiple wavelengths that are the same as the resonance wavelengths of the microrings and maintaining the same modulation response across the entire wavelength range enables wavelength division multiplexing, improves the parallel processing ability of the matrix operation unit, and realizes the expansion of the matrix operation scale.

[0026] In the present invention, by increasing the modulation rate of the microring modulator, modulation sub-signals with multiple symbol periods can exist within the microring, expanding the scale of the multiplication matrix to the corresponding multiple. Without changing the number of device unit structures, arbitrary-scale matrix multiplication operations can be realized by means of time-division multiplexing, expanding the matrix operation scale, eliminating the need for secondary growth of components, reducing manufacturing costs, and enabling flexible regulation.

[0027] In the present invention, multiple microring modulators can be set, allowing real-time reprogramming and changing of the multiplication matrix. It can be combined with the linear operation of a neural network, accelerating the neural network training process, and is very suitable for application scenarios that require online training.

[0028] Further, in step S2, the input matrix X generates a time series arranged in row order x , and the time series x is serially modulated by an external intensity modulator on the amplitude of the optical carrier to obtain an optical signal carrying intensity information.

[0029] Further, the symbol period is equal to one symbol period of the loop recursive delay , that is .

[0030] Further, in step S4, the multiplication matrix W is sorted by column; in order to obtain the multiplication and addition operation results of each row and column of the input matrix X and the multiplication matrix W, the elements need to be traversed. Therefore, a column vector of the multiplication matrix W needs to be repeatedly loaded until the modulation of a time series x ends; when the next time series x is input again, the next column vector of the multiplication matrix W is continuously and repeatedly loaded until X all the row-column multiplications and additions of

[0031] and W are completed. Further, in step S4, the multiplication matrix W is re-edited and converted into to form a second linearly arranged sequence and loaded on the microring modulator, that is:

[0032]

[0033] In the formula, is the coupling matrix of the input microring modulator, which is modulated on the optical intensity signal by a Mach-Zehnder electro-optic modulator; is the element in the n-th row and i-th column of the multiplication matrix W, is the coupling matrix The element in the n-th row and i-th column, is the element in the last row and i-th column of the multiplication matrix W, is The element in the subsequent row of the same column in the coupling matrix

[0034] Furthermore, the optical intensity signal is input into the microring modulator through the first input port of the microring modulator. The multiplication operation between the input matrix and the multiplication matrix is realized by controlling that the symbol period of the first linear sequence in the time domain is equal to that of the second linear sequence loaded on the microring modulator; the newly generated optical signal at the first output port of the microring modulator returns along the microring waveguide to the second input port, where the ring recursive delay matches the symbol period, and just injects the next input symbol at the first input port; then, it is accumulated through coherent optical superposition. Finally, the amplitude of the modulation signal in the microring modulator satisfies:

[0035]

[0036] In the formula, is the element in the i -th row and j -th column of the final result Y, is the element in the last column of the i-th row of the input matrix X, is the coupling matrix The element in the last row and j-th column, is the element in the i-th row and -th column of the input matrix X, is the coupling matrix The -th row and j-th column element, is The element in the subsequent row of the same column in the coupling matrix is the element in the m-th row and j-th column of the multiplication matrix W.

[0037] Finally, the result of the -th row and i -th column is obtained, which also corresponds to the vector dot product result of the j -th row of the input matrix X and the i -th column of the multiplication matrix. The result j is obtained through a photodetector. . ​​

[0038] Further, the signal amplitude-encoded in the microring modulator is imported into the second output port, and a photodetector is used to read the result of the current operation from the second output port.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] A multi-dimensional expandable matrix operation system and its control method of the present invention improve the parallel processing ability of matrix operations and realize the expansion of the scale of matrix operations; without changing the number of device unit structures, any-scale matrix multiplication operations can be realized by time-division multiplexing, expanding the scale of matrix operations, without secondary growth of components, reducing the manufacturing cost, and realizing flexible control; in addition, by setting multiple microring modulators, real-time reprogramming and changing the multiplication matrix are allowed, and it can be combined with neural network linear operations, accelerating the neural network training process, and is very suitable for application scenarios that require online training; moreover, the system of the present invention can be integrated on a chip, which can be integrated to expand the computing parallelism or throughput, so as to obtain a higher data volume. Description of the Drawings

[0041] Figure 1 is a schematic structural diagram of a multi-dimensional expandable matrix operation system of the invention;

[0042] Figure 2 is a schematic structural diagram of the microring modulator of the present invention;

[0043] Figure 3 is a schematic calculation diagram of the matrix operation system in Embodiment 3;

[0044] Figure 4 is a schematic calculation diagram of the matrix operation system in Embodiment 4;

[0045] Figure 5 is a schematic calculation diagram of the matrix operation system in Embodiment 5;

[0046] Figure 6 is an experimental result diagram of extracting the image edge by using the multi-dimensional expandable matrix operation system in Embodiment 6.

[0047] In the drawings: 100, tunable laser; 200, intensity modulator; 300, wavelength division multiplexer; 400, microring modulator; 401, first multimode interferometer; 402, second multimode interferometer; 403, Mach-Zehnder electro-optic modulator; 404, first input port; 405, second input port; 406, first output port; 407, second output port; 500, demultiplexer; 600, photodetector. Detailed Embodiments

[0048] The present invention will be further described below in conjunction with specific embodiments. Among them, the attached drawings are only for illustrative purposes, showing only schematic diagrams, not physical diagrams, and should not be construed as limiting the present invention; in order to better illustrate the embodiments of the present invention, some components in the attached drawings will be omitted, enlarged or reduced, which does not represent the size of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the attached drawings may be omitted.

[0049] In the attached drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the attached drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the attached drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0050] Embodiment 1

[0051] This embodiment is an embodiment of a multi-dimensional expandable matrix operation system, as Figure 1 shown, including:

[0052] Tunable laser 100: used to generate optical carrier signals with multiple carrier wavelengths;

[0053] Intensity modulator 200: used to modulate the optical carrier signal according to the first linear sequence in the time domain, load the optical carrier signal on the optical intensity, and obtain an optical intensity signal carrying intensity information;

[0054] Wavelength division multiplexer 300: used to multiplex and input optical intensity signals with different wavelengths into micro-ring modulator 400;

[0055] Micro-ring modulator 400: used to modulate the optical intensity signal according to the second linear sequence in the time domain to obtain a modulated signal; and superimpose the modulated signals with different symbol periods through micro-ring feedback to obtain an output optical signal; the output optical signal carries the matrix multiplication and addition operation result of the first linear sequence and the second linear sequence;

[0056] Demultiplexer 500: used to separate output optical signals with different wavelengths onto each channel;

[0057] Photodetector 600: used to detect the intensity of the output optical signals on each channel.

[0058] In this embodiment, the first linear sequence is a time sequence formed by arranging the input matrices in the corresponding multiplication order; multiple input matrices simultaneously form multiple parallel time sequences in the same order, that is, multiple first linear sequences; the multiple first linear sequences are encoded by multiple intensity modulators 200 onto different optical carrier wavelengths; the second linear sequence is a time sequence formed by arranging the multiplication matrices in the multiplication order corresponding to the input matrices.

[0059] In this embodiment, the input matrices are arranged in row order to form the first linear sequence; the multiplication matrices are arranged in column order to form the second linear sequence.

[0060] In this embodiment, as Figure 2 shown, the microring modulator 400 includes a 2×2 first multimode interferometer 401 and a second multimode interferometer 402, and a Mach-Zehnder electro-optic modulator 403. The first multimode interferometer 401, the Mach-Zehnder electro-optic modulator 403, and the second multimode interferometer 402 are cascaded, and the first input port 404 of the first multimode interferometer 401 is connected to the first output port 406 of the second multimode interferometer 402 to form a microring; the second input port 405 of the first multimode interferometer 401 is connected to the output end of the wavelength division multiplexer 300, and the second output port 407 of the second multimode interferometer 402 is connected to the input port of the photodetector 600; the arm lengths of the Mach-Zehnder interferometer are controlled by regulating electrodes for intensity modulation.

[0061] In this embodiment, the microring modulator 400 is integrated on a thin-film lithium niobate substrate. In the present invention, the scalable matrix operation system can be integrated on a chip, enabling integration to expand the computational parallelism or throughput, thereby obtaining a higher data volume.

[0062] In this embodiment, the intensity modulator 200 is the Mach-Zehnder electro-optic modulator 403.

[0063] A multi-dimensional scalable matrix operation system provided in this embodiment loads the input matrices on light waves of different wavelengths, utilizing the wavelength dimension of light to expand the scale number of matrix operations. Inputting the wavelength-division multiplexed optical intensity signals onto different microring modulators 400 utilizes spatial division multiplexing to demonstrate the scalability of the matrix operation unit. Converting the matrix information into a time sequence and using time division multiplexing improves the bandwidth utilization rate and also expands the matrix operation scale. At the same time, it supports real-time training and flexibly regulates the information of each matrix element. The present invention can flexibly control the input of matrix operations using various dimensions of light, can accelerate data processing in parallel, and can quickly solve large-scale problems.

[0064] This embodiment provides a compact and flexible photon matrix processing system with powerful multi-dimensional parallel scalability, capable of expanding matrix operations in three dimensions of light, demonstrating outstanding capabilities in high computational throughput and operational versatility using time-division multiplexing, and by further integrating strategies of wavelength-division multiplexing and space-division multiplexing, it can process higher data volumes and exhibit powerful parallel processing capabilities.

[0065] Embodiment 2

[0066] This embodiment is the first embodiment of a control method for a multi-dimensional expandable matrix operation system. This embodiment utilizes the system provided in Embodiment 1 and specifically includes the following steps:

[0067] The present invention also provides a regulation method for a multi-dimensional expandable matrix operation system, including the following steps:

[0068] S1. Use a tunable laser 100 to output multiple optical carrier signals with different wavelengths;

[0069] S2. Arrange the input matrix in a time series in the time domain according to the corresponding multiplication sequence to obtain a first linear sequence, and use an intensity modulator 200 to encode the first linear sequence on optical signals with different wavelengths to obtain an optical intensity signal carrying intensity information;

[0070] S3. Load the optical intensity signals on different wavelength channels onto the same channel through a wavelength-division multiplexer 300 and input them into different microring modulators 400;

[0071] S4. Arrange the multiplication matrix in a time series in the time domain according to the multiplication order corresponding to the input matrix to obtain a second linear sequence; modulate the microring modulator 400 according to the second linear sequence and in the manner of time-division multiplexing to obtain a modulated signal; use the delay of the microring to control the superposition of the modulated signals with different symbol periods to obtain an output optical signal, and control the output optical signal to output to the wavelength-division multiplexer 300;

[0072] S5. Separate the output optical signals with different wavelengths to each channel through a demultiplexer 500 and control the output of a photodetector 600.

[0073] In this embodiment, the process of matrix operation includes:

[0074] In step S2, the input matrix X is arranged in row order to generate a time series x, and the time series x is serially modulated on the amplitude of the optical carrier through an external intensity modulator 200 to obtain an optical signal carrying intensity information.

[0075] The symbol period is equal to one symbol period of the ring recursive delay , that is .

[0076] In step S4, the multiplication matrix W is sorted by column; in order to obtain the multiplication and addition operation results of each row and each column of the input matrix X and the multiplication matrix W, it is necessary to traverse the elements, so it is necessary to repeatedly load a column vector of the multiplication matrix W until a time series x the modulation ends; when the next time series x is input again, continue to repeatedly load the next column vector of the multiplication matrix W until X all the multiplication and addition of rows and columns of and W are completed.

[0077] Since the signal accumulated in the microring is continuously affected by the newly modulated and loaded data of the microring modulator 400, in order to ensure that the finally superimposed and read optical intensity signal is the multiplication of the corresponding row and column elements, the multiplication matrix W needs to be re-edited and converted into a second linear arrangement sequence is formed and loaded on the microring modulator 400, that is:

[0078]

[0079] In the formula, is the coupling matrix input to the microring modulator, and is modulated on the optical intensity signal through the Mach-Zehnder electro-optic modulator; is the element in the nth row and ith column of the multiplication matrix W, is the coupling matrix the element in the nth row and ith column, is the element in the last row and ith column of the multiplication matrix W, is the element in the subsequent row of the same column in the coupling matrix .

[0080] Among them, the optical intensity signal is input to the microring modulator 400 through the first input port 404 of the microring modulator 400, and the multiplication operation of the input matrix and the multiplication matrix is realized by controlling that the code element periods of the first linear sequence in the time domain and the second linear sequence loaded on the microring modulator 400 are equal; the newly generated optical signal at the first output port 406 of the microring modulator 400 returns to the second input port 405 along the microring waveguide, where the ring recursive delay matches the code element period, and just injects the next input symbol at the first input port 404; then it is accumulated through coherent optical superposition (i.e., amplitude summation), and finally the amplitude of the modulation signal in the microring modulator 400 satisfies:

[0081]

[0082] In the formula, is the i row of the final result Y jThe elements of the column, is the element in the last column of the i-th row of the input matrix X, The coupling matrix The element in the last row and column j, is the i-th row of the input matrix X. Column elements, The coupling matrix No. The element in row and column j, for In the coupling matrix Elements in subsequent rows of the same column, is the element in the mth row and jth column of the multiplication matrix W.

[0083] Finally got No. i Line j The results of the columns also correspond to the input matrix X No. i Row and multiply matrix j The vector dot product of the column is obtained by the photodetector 600. .

[0084] To get the result It does not affect the next y value solution, and the signal encoded with amplitude in the micro-ring modulator 400 The result of the current operation is imported into the second output port 407, and the photoelectric detector is used to read the result of the current operation from the second output port 407.

[0085] In this embodiment, the multiple wavelengths output by the tunable laser 100 are controlled to be the same as the resonant wavelength of the microring, and the multiple carrier wavelengths are aligned with their optical resonance through the precise dispersion management of the Mach-Zehnder electro-optic modulator 403 and the ring waveguide in the microring modulator 400 to maintain the same modulation response over the entire wavelength. The tunable laser 100 is controlled to output multiple wavelengths that are the same as the resonant wavelength of the microring, and the same modulation response is maintained over the entire wavelength, which realizes wavelength division multiplexing, improves the parallel processing capability of the matrix operation unit, and realizes the expansion of the scale of matrix operation.

[0086] In this embodiment, by increasing the modulation rate of the micro-ring modulator 400, the micro-ring can have multiple modulation sub-signals of code element periods, so that the scale of the multiplication matrix can be expanded to a corresponding multiple. It is possible to achieve matrix multiplication operations of any scale by time division multiplexing without changing the number of device unit structures, expand the scale of matrix operations, eliminate the need for secondary growth of components, reduce manufacturing costs, and achieve flexible regulation.

[0087] In this embodiment, multiple microring modulators 400 can also be provided, allowing real-time reprogramming and changing of the multiplication matrix. It can be combined with the linear operation of the neural network, accelerating the neural network training process, and is very suitable for application scenarios that require online training.

[0088] Embodiment III

[0089] This embodiment is the second embodiment of a control method for a multi-dimensional expandable matrix operation system. This embodiment is similar to Embodiment II, except that in this embodiment, the control of the scale-expanded matrix operation system is achieved through time-division multiplexing of the microring modulator 400, as Figure 3 shown.

[0090] In this embodiment, by increasing the modulation rate of the microring modulator 400 by N times, the symbol period of the second linear sequence is made , and in the drawings, the microring modulation rate is increased by 2 times.

[0091] In this embodiment, the second linear sequence is loaded onto the electrical input of the microring modulator 400. The microring modulator 400 multiplies it with the input matrix and accumulates the optical signal using its feedback loop. By increasing the modulation rate of the microring modulator 400, time-division multiplexing can be performed on the input of the second linear sequence.

[0092] In this embodiment, the symbol period of the second linear sequence is shortened, enabling the first linear sequence to multiply with multiple data within one symbol, reducing the number of times the first linear sequence is repeatedly input, and shortening the time taken for multiplying two matrices.

[0093] In this embodiment, multiple multiplication matrices can also be cross-input to form a second linear arrangement sequence containing multiple matrix information, achieving the scale expansion of matrix operations.

[0094] Embodiment IV

[0095] This embodiment is the third embodiment of a control method for a multi-dimensional expandable matrix operation system. This embodiment is similar to Embodiment II, except that in this embodiment, the control of the scale-expanded matrix operation system is achieved through wavelength-division multiplexing, as Figure 4 shown.

[0096] In this embodiment, the microring modulator 400 can be compatible with wavelength-division multiplexing. Through precise dispersion management of the Mach-Zehnder interferometer and the microring waveguide, multiple carrier wavelengths are aligned with its optical resonance to maintain the same modulation response across the entire wavelength. The so-called precise dispersion management means that the symbol periods are the same.

[0097] The tunable laser 100 inputs M different wavelengths, serializes the data of each row of the input matrix to form different first linear sequences, loads them on different wavelengths, combines the different wavelength channels and inputs them into the microring modulator 400 to perform matrix operations in parallel. The output ports of the microring modulator 400 are divided into different wavelength channels, and the photodetector 600 is used to read out the optical signals of each wavelength. In this way, the matrix operation rate can be increased by M times.

[0098] Before combining different wavelength channels, multiple input matrices can be serialized to form multiple time sequences, and the multiple time sequences are encoded onto different optical carrier wavelengths using the intensity modulator 200. Combine different wavelength channels and input them into the microring modulator 400 to perform multiple input matrix operations in parallel, realizing the expansion of the matrix scale.

[0099] Embodiment Five

[0100] This embodiment is the fourth embodiment of a control method for a multi-dimensional expandable matrix operation system. This embodiment is similar to Embodiment Two, the difference being that in this embodiment, the control of the matrix operation system with scale expansion is realized by using multi-dimensional parallelism, as Figure 5 shown.

[0101] In this embodiment, the ring modulator can be compatible with wavelength division multiplexing. Through precise dispersion management of the Mach-Zehnder interferometer and the microring waveguide (microring modulator 400), multiple carrier wavelengths are aligned with their optical resonances to maintain the same modulation response across the wavelengths.

[0102] In addition, the tunable laser 100 inputs M different wavelengths, serializes the data of the input matrix to form different first linear sequences and loads them on different wavelengths. Further combining spatial parallelism, an array of K microring modulators 400 is placed after combining different wavelength channels and shares the combined input data.

[0103] The data of each column of the multiplying matrices is loaded in parallel on the second linear sequence of the ring modulator, and two matrix multiplication operations can be completed within one symbol period.

[0104] At the same time, M different wavelengths are used, M input matrices are serialized to form a time sequence, loaded onto different optical carriers using the intensity modulator 200, and further combining spatial parallelism, an array of K microring modulators 400 is placed after combining different wavelength channels and inputs are demultiplexed. Serialize the K multiplying matrices to form a sequence and load it onto the array of microring modulators 400, and the computing throughput will increase from B using only time division multiplexing to BMK times by integrating wavelength division multiplexing and space division multiplexing, where B is the modulation symbol rate.

[0105] Embodiment Six

[0106] In this embodiment, as Figure 6 shown, edge extraction is performed by using two ring modulators and two wavelengths. The extracted results are input into the fully electrical connection layer of a computer to construct a neural network for handwritten digit classification experiments. The experiments verify the parallel scalability of the proposed scheme. The TFLN microring modulator used in the experiments has a bandwidth exceeding 67 GHz, an area of 3.4 mm × 0.7 mm, and a free spectral range (FSR) of 0.148 nm. The selected wavelengths correspond to the optical resonances of the microring modulators. For optical edge extraction, 300 handwritten digit images were selected, each with a resolution of 28 × 28 pixels, as Figure 6 shown in (a) of Figure 6 . Due to the limitations of the available test equipment, all 300 images were sequentially loaded onto the first wavelength, convolved with the first kernel on the microring modulator, and then loaded onto the next wavelength for convolution with the second kernel on the same microring modulator, as Figure 6 shown in (b) of Figure 6 . After convolution, the results of edge extraction are respectively as Figure 6 shown in (c) of Figure 6 . During the experiment, the operating conditions of the microring modulator - including the bias voltage, temperature, and electro-optic spectral response - remained unchanged for the input data of the two wavelengths. According to the confusion matrix in (d) of Figure 6 , the final classification accuracy was ~87%, demonstrating the feasibility of a multi-dimensional scalable matrix operation unit and its control method.

[0107] In the specific content of the above specific implementation manner, the technical features can be combined arbitrarily without contradiction. For the sake of brevity of description, not all possible combinations of the above technical features are described. However, as long as the combinations of these technical features do not conflict, they should all be considered to be within the scope described in this specification.

[0108] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention and are not intended to limit the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention should all be included within the protection scope of the claims of the present invention.

Claims

1. A multi-dimensional scalable matrix operation system, characterized in that: include: Tunable laser (100): used to generate optical carrier signals of multiple carrier wavelengths; An intensity modulator (200): used to modulate an optical carrier signal according to a first linear sequence in the time domain, load the optical carrier signal onto the light intensity, and obtain an optical intensity signal carrying intensity information; A wavelength division multiplexer (300): used for multiplexing light intensity signals of different wavelengths and inputting them into a micro-ring modulator (400); A micro-ring modulator (400) is used to modulate an optical intensity signal according to a second linear sequence in the time domain to obtain a modulated signal; and to superimpose modulated signals of different symbol periods through micro-ring feedback to obtain an output optical signal; the output optical signal carries a matrix multiplication and addition result of a first linear sequence and a second linear sequence; the first linear sequence is a time sequence formed by arranging input matrices in a corresponding multiplication order; a plurality of input matrices simultaneously form a plurality of parallel time sequences in the same order, i.e., a plurality of first linear sequences; a plurality of the first linear sequences are encoded by a plurality of the intensity modulators (200) onto different optical carrier wavelengths; the second linear sequence is a time sequence formed by arranging multiplication matrices in a multiplication order corresponding to the input matrices; De-Wavelength Division Multiplexer (500): used to separate output optical signals of different wavelengths into various channels; Photoelectric detector (600): used to detect the intensity of the output optical signal on each channel; The optical intensity signal is input into the microring modulator (400) through the first input port (404) of the microring modulator (400), and the multiplication operation of the input matrix and the multiplication matrix is ​​realized by controlling the symbol period of the first linear sequence in the time domain to be equal to the symbol period of the second linear sequence loaded on the microring modulator (400); the newly generated optical signal at the first output port (406) of the microring modulator (400) returns to the second input port (405) along the microring waveguide, wherein the microring recursive delay matches the symbol period, and the next input symbol is injected at the first input port (404); and then accumulation is performed through coherent optical superposition, and finally the amplitude of the modulated signal in the microring modulator (400) satisfies: In the formula, is the final result Y i Line j The elements of the column, is the element in the last column of the i-th row of the input matrix X, The coupling matrix The last row and column j element, is the i-th row of the input matrix X. Column elements, The coupling matrix No. The element in row and column j, for In the coupling matrix Elements in subsequent rows of the same column, is the mth row and jth column element of the multiplication matrix W; Finally get No. i Line j The results of the columns also correspond to the input matrix X No. i Row and multiply matrix j The vector dot product of the columns is obtained by the photodetector (600). .

2. The multi-dimensional scalable matrix operation system according to claim 1, characterized in that: The micro-ring modulator (400) comprises a 2×2 first multi-mode interferometer (401) and a second multi-mode interferometer (402), and a Mach-Zehnder electro-optic modulator (403); the first multi-mode interferometer (401), the Mach-Zehnder electro-optic modulator (403), and the second multi-mode interferometer (402) are cascaded, and the first input port (404) of the first multi-mode interferometer (401) is connected to the first output port (406) of the second multi-mode interferometer (402) to form a micro-ring; the second input port (405) of the first multi-mode interferometer (401) is connected to the output end of the wavelength division multiplexer (300), and the second output port (407) of the second multi-mode interferometer (402) is connected to the input port of the photodetector (600); and the arm length of the Mach-Zehnder interferometer is controlled by regulating electrodes to perform intensity modulation.

3. A control method for a multi-dimensional scalable matrix operation system according to any one of claims 1 or 2, characterized in that: The following steps are involved: S1. Using a tunable laser (100) to output a plurality of optical carrier signals of different wavelengths; S2. Arranging the input matrix into a time series in the time domain according to the corresponding multiplication sequence to obtain a first linear sequence, and encoding the first linear sequence on optical signals of different wavelengths using an intensity modulator (200) to obtain an optical intensity signal carrying intensity information; S3. The optical intensity signals on different wavelength channels are loaded on the same channel through the wavelength division multiplexer (300) and input to different micro-ring modulators (400); S4. Arrange the multiplication matrix in a multiplication order corresponding to the input matrix to form a time sequence in the time domain, and obtain a second linear sequence; modulate the micro-ring modulator (400) according to the second linear sequence and in a time division multiplexing manner to obtain a modulated signal; use the delay of the micro-ring to control the superposition of modulated signals of different symbol periods to obtain an output optical signal, and control the output optical signal to be output to the wavelength division multiplexer (300); S5. The output optical signals of different wavelengths are separated into various channels through the wavelength division multiplexer (500), and the output of the photodetector (600) is controlled.

4. The control method of the multi-dimensional scalable matrix operation system according to claim 3, characterized in that: In step S2, the input matrix X Generate time series by row order x , time series x The amplitude of the optical carrier is serially modulated by an external intensity modulator (200) to obtain an optical signal carrying intensity information.

5. The control method of the multi-dimensional scalable matrix operation system according to claim 4, characterized in that: The symbol period is equal to one symbol period of the microloop recursive delay ,Right now .

6. The control method of the multi-dimensional scalable matrix operation system according to claim 5, characterized in that: In step S4, the matrix W is multiplied Sort by columns; to get the input matrix X The multiplication and addition results of each row and column of the multiplication matrix W are traversed, and a column vector of the multiplication matrix W is repeatedly loaded until a time series x Modulation ends; when the next time series x When input again, continue to repeatedly load the next column vector of the multiplication matrix W until X All row and column multiplication and addition of W are completed.

7. The control method of the multi-dimensional scalable matrix operation system according to claim 6, characterized in that: In step S4, the multiplication matrix W is re-edited and converted into A second linear arrangement sequence is formed and loaded on the micro-ring modulator (400), namely: In the formula, The coupling matrix of the input micro-ring modulator is modulated on the optical intensity signal through the Mach-Zehnder electro-optic modulator; is the nth row and ith column element of the multiplication matrix W, The coupling matrix The element at row n and column i, is the last row and column i element of the multiplication matrix W, for In the coupling matrix Elements in subsequent rows of the same column.

8. The control method of the multi-dimensional scalable matrix operation system according to claim 7, characterized in that: The signal encoded with amplitude in the micro-ring modulator (400) The result of the current operation is imported into the second output port (407) and read from the second output port (407) using the photoelectric detector (600).

Images