Optical computing device and optical signal processing method

Abstract

An optical computing device and an optical signal processing method are disclosed. The optical computing device includes a parametric oscillator array, an interaction computing matrix connected to the parametric oscillator array, a first feedback module connected to both ends of the parametric oscillator array, and a second feedback module connected between the parametric oscillator array and the interaction computing array. The parametric oscillator array is configured to receive a first group of signals and generate a first group of optical signals comprising a plurality of first optical signals based on the received first group of signals; the interaction computing array is configured to receive the first group of optical signals and perform matrix operations on the first group of optical signals according to a preset matrix to obtain a second group of optical signals comprising a plurality of second optical signals; the first feedback module is configured to receive the first group of optical signals and transmit the first group of optical signals to the parametric oscillator array; and the second feedback module is configured to receive the second group of optical signals and transmit the second group of optical signals to the parametric oscillator array.

Description

Technical Field

[0001] The present application relates to the field of information technology, and in particular to an optical computing device and an optical signal processing method. Background Art

[0002] Non-deterministic polynomial hard (NP-hard) problems, such as the analysis and optimization of complex systems, can be mapped into an Ising model, which can be simulated and solved using an optical Ising machine to obtain the solution to the NP-hard problem.

[0003] An optical Ising machine uses a lattice network composed of multiple optical signals to simulate and solve the Ising model. Currently, one implementation of an optical Ising machine involves a fiber optic system. Specifically, all the input optical signals required for the optical Ising machine calculation are serially fed into the machine. These signals are then circulated within the fiber cavity. A small portion of the input optical signal is coupled out and used in a field-programmable gate array (FPGA) for matrix operations, simulating the coupling between the nodes in the Ising model. The result of the operation is then fed back to the input of the optical Ising machine, where it interacts with all the input optical signals and serves as the next input optical signal. This process is repeated to ultimately obtain the optimal solution for the optical Ising machine.

[0004] Because the optical Ising machine incorporates all the input optical signals required for computation, the optimal solution obtained by the machine is a global, rather than a local, optimal solution, thus ensuring the machine's computational accuracy. However, in the optical Ising machine, the input optical signals are serially transmitted. When the Ising model has a large number of nodes, a large number of input optical signals are required, which increases the transmission time of the input optical signals within the machine, significantly increasing computation time and limiting the machine's computational efficiency.

[0005] Therefore, how to improve the computing efficiency of the optical Ising machine while ensuring calculation accuracy is an urgent problem to be solved. Summary of the Invention

[0006] The present application provides an optical computing device and an optical signal processing method, so as to provide an optical Ising machine with high computing accuracy and high computing efficiency.

[0007] In a first aspect, an optical computing device is provided. The optical computing device includes a parametric oscillator array, an interaction computing matrix connected to the parametric oscillator array, a first feedback module connected to both ends of the parametric oscillator array, and a second feedback module connecting the parametric oscillator array and the interaction computing matrix. The operating principle of the optical computing device is as follows:

[0008] A parametric oscillator array is configured to receive a first group of signals and generate a first group of optical signals including a plurality of first optical signals based on the received first group of signals; an interaction calculation array is configured to receive the first group of optical signals and perform a matrix operation on the first group of optical signals according to a preset matrix to obtain a second group of optical signals including a plurality of second optical signals; a first feedback module is configured to receive the first group of optical signals and transmit the first group of optical signals to the parametric oscillator array; and a second feedback module is configured to receive the second group of optical signals and transmit the second group of optical signals to the parametric oscillator array.

[0009] In the aforementioned optical computing device, the parametric oscillator array and the interaction computation matrix process multiple optical signals included in a group of optical signals in parallel. Therefore, even when the number of nodes in the Ising model is large and more input optical signals need to be introduced, the operation time of the optical computing device will not be increased.

[0010] Furthermore, the optical signal generated by the parametric oscillator array is fed back to the parametric oscillator array for the next calculation, which can increase the state of the optical signal generated by the parametric oscillator array, so that the optical computing device can obtain input optical signals of all possible states required for calculation, thereby avoiding the optical computing device obtaining a non-local optimal solution, that is, using hardware annealing to improve the calculation accuracy of the optical computing device.

[0011] In one possible design, the parametric oscillator array is further configured to receive the first and second optical signals, and output a third optical signal based on the first, first, and second optical signals. The interaction computation array is further configured to receive the third optical signal and perform matrix operations on the third optical signal according to the preset matrix to obtain a fourth optical signal. The first feedback module is further configured to receive the third optical signal and transmit it to the parametric oscillator array. The second feedback module is further configured to receive the fourth optical signal and transmit it to the parametric oscillator array.

[0012] In the above optical computing device, after the parametric oscillator array receives the optical signals transmitted by the first feedback module and the second feedback module, it can combine the first group of signals and the optical signals transmitted by the first feedback module and the second feedback module to generate an input signal for calculation.

[0013] In one possible design, the parametric oscillator array includes a plurality of parametric oscillators, each of which includes a first waveguide and a second waveguide. The first waveguide and the second waveguide are made of different materials, and the material of the first waveguide includes a material having a nonlinear effect.

[0014] In the aforementioned optical computing device, the parametric oscillators in the parametric oscillator array can be integrated via waveguides, making the optical computing device more compact. Furthermore, the first waveguide is a waveguide with nonlinear effects, such as a lithium niobate waveguide or a lithium tantalate waveguide with second-order nonlinearity. The second waveguide can be a waveguide with low transmission loss, such as a silicon nitride waveguide, or silicon, silicon dioxide, or the like. The nonlinear effects of the first waveguide can generate a compressed optical signal, which can be understood as the input optical signal in all possible states required for the calculation. The generated compressed optical signal is then transmitted via the second waveguide with low transmission loss, thereby preventing the optical computing device from obtaining a non-locally optimal solution. By combining the advantages of different waveguide materials through heterogeneous integration, the computational accuracy of the optical computing device can be guaranteed.

[0015] In one possible design, the material of the second waveguide includes a material with a transmission loss below a threshold.

[0016] In this way, the loss of the optical signal during transmission in the second waveguide can be reduced, thereby ensuring the accuracy of the optical signal.

[0017] In one possible design, the interaction computing array includes multiple cascaded Mach-Zehnder (MZ) interferometers, each of which includes multiple Mach-Zehnder interferometers (MZIs) and beam splitters. Each MZI includes a waveguide with a dielectric constant adjustment speed less than a threshold, which is used to adjust the phase parameters of the corresponding MZI.

[0018] The small size of the Mach-Zehnder interferometer (MZI) unit allows for compact optical computing devices. Furthermore, this allows for on-chip implementation, ensuring system stability. Furthermore, using materials with a dielectric constant adjustment speed slower than the threshold characteristic to form the MZI allows for rapid loading and refreshing of phase parameters, thereby ensuring the computational efficiency of the optical computing device.

[0019] In a possible design, the waveguides in the multiple MZIs included in the same MZ interference unit are made of different materials.

[0020] The MZI formed by heterogeneous integration can quickly load and refresh its phase parameters, thereby enabling the rapid loading and refreshing of the interaction matrix, thereby further improving the computational efficiency of optical computing devices.

[0021] In one possible design, the parametric oscillator further includes: a splitter connected to the parametric oscillator, the first feedback module, and the interaction calculation array, the splitter being configured to: receive an optical signal from the first group of optical signals, and, according to a preset splitting ratio, split the optical signal into a first portion of the optical signal and a second portion of the optical signal, and transmit the first portion of the optical signal to the first feedback module, and transmit the second portion of the optical signal to the interaction calculation matrix.

[0022] The optical splitter can flexibly control the energy of the optical signal transmitted to the first feedback module and the interaction calculation matrix, thereby increasing the flexibility of the system.

[0023] In one possible design, the parametric oscillator further includes a combiner connected to the parametric oscillator, the first feedback module, and the second feedback module, wherein the combiner is configured to combine an optical signal of the first group of optical signals and an optical signal of the second group of optical signals into one optical signal, and transmit the combined optical signal to the parametric oscillator.

[0024] By combining multiple signals into one signal through a combiner and then processing it, the amount of calculation of the parametric oscillator can be reduced.

[0025] In the second aspect, the present application provides a method for processing optical signals. The beneficial effects can be found in the relevant description of the first aspect and will not be repeated here. The method is performed by an optical computing device, which includes a parametric oscillator array, an interaction computing array, a first feedback module, and a second feedback module. The method includes: the parametric oscillator array receives a first group of signals, and generates a first group of optical signals based on the first group of signals, the first group of optical signals including a plurality of first optical signals; the interaction computing array receives the first group of optical signals, and performs a matrix operation on the first group of optical signals according to a preset matrix to obtain a second group of optical signals, the second group of optical signals including a plurality of second optical signals; the first feedback module receives the first group of optical signals and transmits the first group of optical signals to the parametric oscillator array; the second feedback module receives the second group of optical signals and transmits the second group of optical signals to the parametric oscillator array.

[0026] In one possible design, the parametric oscillator array receives the first group of optical signals and the second group of optical signals, and outputs a third group of optical signals based on the first group of signals, the first group of optical signals, and the second group of optical signals; the interaction computation array receives the third group of optical signals, performs matrix operations on the third group of optical signals according to the preset matrix, and obtains a fourth group of optical signals; the first feedback module receives the third group of optical signals and transmits the third group of optical signals to the parametric oscillator array; and the second feedback module receives the fourth group of optical signals and transmits the fourth group of optical signals to the parametric oscillator array.

[0027] In one possible design, the parametric oscillator array includes multiple parametric oscillators, each of which receives a first optical signal in the first group of optical signals and splits the first optical signal into a first partial optical signal and a second partial optical signal according to a preset splitting ratio; each parametric oscillator transmits the first partial optical signal to the first feedback module, and transmits the second partial optical signal to the interaction calculation matrix.

[0028] In one possible design, the parametric oscillator array includes multiple parametric oscillators, each parametric oscillator combines a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmits the combined optical signal to the parametric oscillator.

[0029] In a third aspect, the present application provides an optical computing chip, which may include the optical computing device described in the first aspect or any possible implementation of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A schematic diagram of an example structure of an optical computing device 10 provided in an embodiment of the present application;

[0031] Figure 2 A schematic structural diagram of an example of a parametric oscillator array 100 provided in an embodiment of the present application;

[0032] Figure 3 A schematic structural diagram of an example of an optical parametric oscillator provided in an embodiment of the present application;

[0033] Figure 4 A schematic diagram of an example structure of an optical parametric oscillation region provided in an embodiment of the present application;

[0034] Figure 5 A schematic diagram of an example structure of the interaction calculation matrix 200 provided in an embodiment of the present application;

[0035] Figure 6 A schematic diagram of an example structure of an MZIU provided in an embodiment of the present application;

[0036] Figure 7 A schematic structural diagram of another example of an MZIU provided in an embodiment of the present application;

[0037] Figure 8 This is a structural diagram of an example of providing a spectrometer between the parametric oscillator 110 and the interaction calculation matrix 200 in an embodiment of the present application;

[0038] Figure 9 4 is a structural diagram of an example of providing a beam combiner between the parametric oscillator 110 and the second feedback module 400 in an embodiment of the present application;

[0039] Figure 10 A flowchart of a method for processing an optical signal provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0041] In the embodiments of the present application, "multiple" refers to two or more. In view of this, in the embodiments of the present application, "multiple" can also be understood as "at least two". "At least one" can be understood as one or more, for example, one, two or more. For example, including at least one means including one, two or more, and does not limit which ones are included. For example, including at least one of A, B and C, then the included ones may be A, B, C, A and B, A and C, B and C, or A, B and C. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / ", unless otherwise specified, generally indicates that the previous and subsequent associated objects are in an "or" relationship.

[0042] Unless otherwise specified, ordinal numbers such as "first" and "second" in the embodiments of the present application are used to distinguish multiple objects and are not used to limit the order, timing, priority or importance of multiple objects.

[0043] The present application provides an optical computing device and an optical signal processing method, so as to provide an optical Ising machine with high computing accuracy and high computing efficiency.

[0044] Please refer to Figure 1 , is a structural diagram of an example of an optical computing device 10 provided in an embodiment of the present application. Figure 1As shown, the optical computing device 10 includes a parametric oscillator array 100, an interaction computing array 200, a first feedback module 300, and a second feedback module 400. The parametric oscillator array 100 includes a first input terminal and a first output terminal, the interaction computing array 200 includes a second input terminal and a second output terminal, the first output terminal of the parametric oscillator array 100 is connected to the second input terminal of the interaction computing array 200, the second output terminal of the interaction computing array 200 is connected to the first input terminal of the parametric oscillator array 100 via the second feedback module 400, and the first feedback module 300 is connected to both ends of the parametric oscillator array 100.

[0045] The parametric oscillator array 100 is capable of receiving a first group of signals and generating a first group of optical signals according to the received first group of signals. The first group of optical signals includes at least one first optical signal.

[0046] In an embodiment of the present application, the first group of signals may include optical signals, such as optical pulses; alternatively, the first group of signals may be electrical signals, such as electrical pulses. The first group of signals may include one optical signal or one electrical signal, or may include multiple optical signals or multiple electrical signals. The signal type and quantity of the first group of signals are not limited herein. It should be noted that when the first group of signals received by the parametric oscillator array 100 is an electrical signal, the parametric oscillator array 100 may convert the received electrical signal into an optical signal, and then generate the first group of optical signals based on the converted optical signal.

[0047] As an example, the first group of signals can be sent by a pump source. For example, the pump source can periodically generate a group of signals. The multiple optical signals or electrical signals included in the group of signals can be the same or different. The multiple optical signals or electrical signals in the group of signals arrive at the parametric oscillator array 100 simultaneously. The first group of signals can be any group of signals periodically generated by the pump source. Of course, the first group of signals can also be generated by other means, and the source of the first group of signals is not limited here.

[0048] The interaction calculation matrix 200 may be pre-loaded with a preset matrix for calculation. After receiving the first set of optical signals generated by the parametric oscillator array 100, a matrix operation is performed on the first set of optical signals according to the preset matrix. The preset matrix can be used to adjust parameters such as the phase or amplitude of the first set of optical signals to obtain a second set of optical signals.

[0049] It should be noted that the number of optical signals included in the first group of optical signals and the second group of optical signals can be the same. For example, if the first group of optical signals includes N optical signals, the second group of optical signals also includes N optical signals. In this case, the optical signals in the first group of optical signals and the optical signals in the second group of optical signals are in a one-to-one correspondence; alternatively, the number of optical signals included in the first group of optical signals and the second group of optical signals can also be different. For example, the first group of optical signals includes N optical signals, and the second group of optical signals includes M optical signals, where M and N are both positive integers, and M and N are not equal.

[0050] The first feedback module 300 is configured to, after receiving a first set of optical signals generated by the parametric oscillator array 100, transmit the first set of optical signals to a first input terminal of the parametric oscillator array 100. The second feedback module 400 is configured to, after receiving a second set of optical signals generated by the interaction calculation matrix 200, transmit the second set of optical signals to the first input terminal of the parametric oscillator array 100.

[0051] It can be understood that the parametric oscillator array 100 is also used to receive the first and second groups of optical signals. Thus, in subsequent processing, the parametric oscillator array 100 outputs a third group of optical signals based on the first group of signals, the first group of optical signals fed back by the first feedback module 300, and the second group of optical signals fed back by the second feedback module 400. The interaction computing array 200 is also used to receive the third group of optical signals and perform matrix operations on the third group of optical signals according to the preset matrix to obtain a fourth group of optical signals. The first feedback module 300 is also used to receive the third group of optical signals and transmit them to the first input terminal of the parametric oscillator array 100. The second feedback module 400 is also used to receive the fourth group of optical signals and transmit them to the parametric oscillator array 100. The above process is repeated. After multiple processing steps, the optimal solution of the optical computing device 10 is obtained.

[0052] The above content summarizes the functions of the various modules of the optical computing device 10 , and each module will be described in detail below.

[0053] 1. Parametric oscillator array 100.

[0054] The parametric oscillator array 100 includes a plurality of parallel parametric oscillators 110, wherein the number of the plurality of parametric oscillators 110 may be the same as the number of optical signals or electrical signals included in the first group of signals received by the parametric oscillator array 100; or, since the parametric oscillator array 100 is also used to receive the first group of optical signals and the second group of optical signals, the number of the plurality of parametric oscillators 110 may also be the same as the number of optical signals included in the first group of optical signals or the second group of optical signals. If the number of optical signals or electrical signals included in the first group of signals, the number of optical signals included in the first group of optical signals, and the number of optical signals included in the second group of optical signals are different, the number of the plurality of parametric oscillators 110 may be the maximum number of optical signals or electrical signals included in the three groups of signals. For ease of explanation, the following takes the case where the number of optical signals or electrical signals included in the three groups of signals is the same as an example.

[0055] Each of the multiple parallel parametric oscillators 110 is used to receive one of the signals in the first group of signals, and / or to receive one of the optical signals in the first group of optical signals fed back by the first feedback module 300, and / or to receive one of the optical signals in the second group of optical signals fed back by the second feedback module 400.

[0056] Please refer to Figure 2 For example, a parametric oscillator array 100 includes N parallel parametric oscillators 110, and the first group of signals is optical signal group A including optical signals A1, A2, A3 ... AN, the first group of optical signals is optical signal group B including optical signals B1, B2, B3 ... BN, and the second group of optical signals is optical signal group C including optical signals C1, C2, C3 ... CN. For example, the first parametric oscillator 110 is configured to receive optical signals A1, B1, and C1, and to generate an optical signal in a third group of optical signals based on optical signals A1, B1, and C1, labeled as optical signal D1. The second parametric oscillator 110 is configured to receive optical signals A2, B2, and C2, and to generate another optical signal in the third group of optical signals based on optical signals A2, B2, and C2, labeled as optical signal D2, and so on. Since each parametric oscillator 110 can independently process the optical signal it receives, multiple optical signals in the received group of optical signals can be processed in parallel, thereby improving the computational efficiency of the optical computing device 10.

[0057] In an embodiment of the present application, the parametric oscillator 110 may be an optical parametric oscillator, for example, an optically pumped optical parametric oscillator or an electrically pumped optical parametric oscillator cavity 110, may be a laser oscillator or a Kerr oscillator or a polariton, etc., or may be any parametric oscillator that can oscillate based on the frequency of an optical signal, without limitation herein.

[0058] As an example, when the parametric oscillator 110 is an optical parametric oscillator, the optical parametric oscillator may generate a new optical signal through nonlinear optical interaction. Figure 3 , is a schematic diagram of an example structure of an optical parametric oscillator provided in an embodiment of the present application. Figure 3 As shown, the optical parametric oscillator includes two Bragg reflection regions and an optical parametric oscillation region, with the Bragg reflection regions located at both ends of the optical parametric oscillation region. The two Bragg reflection regions form a resonant cavity, and the optical signal received by the optical parametric oscillator is transmitted back and forth between the two Bragg reflection regions, oscillating. The optical signal is coupled into the optical parametric oscillation region, where it undergoes a nonlinear interaction with the optical signals transmitted by the first feedback module 300 and the second feedback module 400, respectively. After the nonlinear interaction, the optical signal is coupled and filtered out of the optical parametric oscillation region. After filtering, the optical signal remaining in the optical parametric oscillator becomes a group of optical signals output by the parametric oscillator array 100, such as the first group of optical signals or the third group of optical signals described above.

[0059] Next, the optical parametric oscillation region in the optical parametric oscillator 110 will be described.

[0060] Please refer to Figure 4 , which is a structural diagram of an example of an optical parametric oscillation region provided in an embodiment of the present application. Figure 4 is the cross-sectional view of the optical parametric oscillation region in the XZ plane of the world coordinate system, as shown Figure 4 As shown, the optical parametric oscillator region includes two waveguides: a first waveguide and a second waveguide. The first waveguide is integrated on the second waveguide and exhibits nonlinear effects. The following describes the operating principles of the first and second waveguides, focusing on the differences in the signals received by the optical parametric oscillator.

[0061] In the first case, the optical parametric oscillator generates the first set of optical signals according to the first set of received signals:

[0062] Taking the first group of signals as pump light as an example, the pump light can enter the first waveguide through end face coupling, obtain the first group of optical signals after the nonlinear effect of the first waveguide, and then vertically couple into the second waveguide.

[0063] In the second case, the optical parametric oscillator generates a third group of optical signals based on the received first group of signals, the first group of optical signals, and the second group of optical signals:

[0064] Continuing with the example of the first group of signals being pump light, the pump light enters the first waveguide through end-face coupling. The first and second groups of optical signals can be vertically coupled into the first waveguide through mode field conversion. Then, the pump light, the first and second groups of optical signals pass through the nonlinear action of the first waveguide together to obtain the third group of optical signals, which are then vertically coupled into the second waveguide.

[0065] It should be noted that in the second case, since the pump light, the first group of optical signals, and the second group of optical signals need to pass through the nonlinear effect of the first waveguide together, the pump light, the first group of optical signals, and the second group of optical signals need to overlap in the time domain when coupled into the first waveguide, and thus it is necessary to control the time at which the pump light, the first group of optical signals, and the second group of optical signals are optically coupled into the first waveguide. As an example, the transmission path length of the optical signal in the first feedback module 300 and the second feedback module 400 and the period of the optical signal generated by the pump source can be adjusted so that the pump light, the first group of optical signals, and the second group of optical signals can all enter the first waveguide simultaneously within a set time period.

[0066] In the embodiments of the present application, the first waveguide and the second waveguide can be waveguides made of different materials. For example, the first waveguide is a waveguide with nonlinear effects, such as a lithium niobate waveguide or a lithium tantalate waveguide with second-order nonlinearity, or a waveguide material with third-order nonlinearity. The second waveguide can be a waveguide with low transmission loss characteristics. A waveguide with low transmission loss characteristics can be understood as a waveguide with transmission loss below a threshold, such as a silicon nitride waveguide, or other semiconductor materials with a wide transmission spectrum and low transmission loss, such as silicon or silicon dioxide. The first waveguide and the second waveguide can be integrated by integrating the first waveguide onto the second waveguide material using heterogeneous integration technology, or by integrating the second waveguide onto the first waveguide material. The specific materials of the first and second waveguides and the integration methods of the two are not limited herein.

[0067] Since a compressed optical signal can be generated in the above-mentioned optical parametric oscillation region through the nonlinear effect of the first waveguide, the compressed optical signal can be understood as an input optical signal of all possible states required for calculation, and the generated compressed optical signal is transmitted through a second waveguide with low transmission loss characteristics, thereby avoiding the optical computing device 10 obtaining a non-local optimal solution. By combining the characteristic advantages of different waveguide materials in a heterogeneous integration manner, the calculation accuracy of the optical computing device 10 can be guaranteed.

[0068] In addition, as can be seen from the above description, the optical signal output by the above-mentioned optical parametric oscillator also includes pump light, which will affect the calculation results of the interaction calculation matrix 200. Therefore, in order to ensure the accuracy of the calculation results of the optical computing device 10, a filter can be set after each parametric oscillator. The filter can be a narrowband microring filter, etc., which filters out the pump light in the optical signal output by the parametric oscillator.

[0069] 2. Interaction calculation matrix 200.

[0070] After the parametric oscillator array 100 outputs the first or third optical signals, the interaction calculation matrix 200 performs a matrix operation on the first or third optical signals, thereby obtaining the corresponding operation results, namely, the second optical signals corresponding to the first optical signals, or the fourth optical signals corresponding to the third optical signals. The processing of the first optical signals by the interaction calculation matrix 200 is the same as that for the third optical signals. The following description uses the processing of the first optical signals by the interaction calculation matrix 200 as an example.

[0071] As an example, the first group of optical signals may be a group of N optical signals, denoted as f(t), which is expressed as follows:

[0072] f(t)=[f1,f2,…,f n ](1)

[0073] Among them, f1,f2,…,f n are the column vectors corresponding to the N optical signals. The interaction calculation matrix 200 constitutes an n*n symmetric matrix, which is related to the NP-hard problem to be solved. For example, the n*n symmetric matrix can be obtained by mathematically abstracting the NP-hard problem. The n*n symmetric matrix is ​​denoted as J and is expressed as follows:

[0074]

[0075] After the first group of optical signals is operated by the interaction calculation matrix 200, an output signal corresponding to the first group of optical signals is obtained, which is recorded as y(t). y(t) satisfies the following expression:

[0076] y(t)=J*f=f1J 11 +f2J 12 +…+f n J 1n (3)

[0077] Please refer to Figure 5, which is a structural diagram of an example of the interaction calculation matrix 200 provided in an embodiment of the present application. The interaction calculation matrix 200 includes a plurality of cascaded Mach-Zehnder interferometer units (MZIU). In the embodiment of the present application, each MZIU can receive two optical signals, and each MZIU is used to realize the interaction between the two input optical signals. The interaction can be the interaction between the phases of the two input optical signals, or the interaction between the signal amplitudes of the two input optical signals, or the interaction between other parameters of the optical signals, which is not limited here. For the convenience of explanation, the interaction is taken as an example of the interaction between the phases of the two input optical signals.

[0078] Each MZIU can control the phase of the two input optical signals through the phase parameters of the MZI, where each element in the aforementioned n*n symmetric matrix corresponds to the phase parameter of each MZIU in multiple cascaded MZIUs, and then the phase parameter of each MZIU is set according to the two-dimensional matrix.

[0079] exist Figure 5 In the interaction calculation matrix 200 shown in FIG. 1 , the optical signal is input from the left end of the interaction calculation matrix 200 and the calculation result of the interaction calculation matrix 200 is output from the right end. Figure 5 As shown, the multiple cascaded MZIUs can be divided into three parts according to the positions shown by the dotted lines in the figure, wherein the first part of the MZIU matrix includes multiple MZIUs in a triangular shape, the second part of the MZIU matrix includes multiple MZIUs in a diagonal cascade, and the third part of the MZIU matrix includes multiple MZIUs in an inverted triangular shape. When setting the phase parameters in each MZIU according to the aforementioned n*n symmetric matrix, the n*n symmetric matrix can be divided into three sub-matrices, and then the three sub-matrices are respectively mapped to the three-part MZIU matrix. Of course, the n*n symmetric matrix can also be mapped to the multiple cascaded MZIUs in other ways, which is not limited here.

[0080] Please refer to Figure 6 , is a schematic diagram of an example structure of the MZIU provided in an embodiment of the present application. Each MZIU includes multiple Mach-Zehnder interferometers (MZIs) and beam splitters. Figure 6In the example, the MZIU includes 2 MZIs (respectively MZI-1 and MZI-2) and 2 beam splitters (respectively beam splitter 1 and beam splitter 2). The lower ports of the two beam splitters are directly connected, and the upper port of each beam splitter is connected to a beam splitter respectively. For example, the upper port of beam splitter 1 is connected to MZI-1, and the upper port of beam splitter 2 is connected to MZI-2. Among them, the splitting ratio of the beam splitter can be set according to actual usage requirements. As an example, the splitting ratio of the beam splitter can be 50 / 50, so that the optical signal transmitted to the MZIU will be evenly transmitted to MZI-1 and the lower beam splitter 2. Moreover, the phase parameters of each MZI can be adjusted by the phase controller of the MZI. The phase parameters of multiple MZIs included in an MZIU can be the same or different. The specific values ​​are determined according to the aforementioned n*n symmetric matrix.

[0081] Please refer to Figure 7 , is a structural diagram of another example of an MZIU provided in an embodiment of the present application. Each MZIU includes two MZIs (MZI-3 and MZI-4) and two beam splitters (beam splitter 3 and beam splitter 4). Figure 7 The connection method between MZI and beam splitter in the MZIU is similar to Figure 6 Instead, the upper ports of the two beam splitters are directly connected, and the lower port of each beam splitter is connected to a beam splitter, for example, the lower port of beam splitter 3 is connected to MZI-3, and the lower port of beam splitter 4 is connected to MZI-4. Figure 6 Similar, no further description is given here.

[0082] As an example, Figure 5 In the multiple cascaded MZIUs shown, the structures of the MZIUs in the first and second MZIU matrices are the same, and the structures of the MZIUs in the third MZIU matrix are different. For example, the MZIUs in the first and second MZIU matrices use Figure 6 The structure of the MZIU shown in the figure, and the MZIU in the third part of the MZIU matrix uses Figure 7 Alternatively, the MZIU matrix in the first part and the MZIU matrix in the second part use Figure 7 The structure of the MZIU shown in the figure, and the MZIU in the third part of the MZIU matrix uses Figure 6 The structure of the MZIU shown is not limited here.

[0083] In addition, in the embodiment of the present application, Figure 6The two MZIs included in the MZIU can be formed from materials with dynamic phase modulation properties. These materials can be materials with electro-optical effects, such as lithium niobate, organic polymers, or materials with magneto-optical effects, all of which are not listed here. Furthermore, the two MZIs in an MZIU can be formed from different materials with dynamic phase modulation properties. For example, one MZI can be formed from lithium niobate and the other from an organic polymer; or one MZI can be formed from a material with magneto-optical effects and the other from an electro-optical effect.

[0084] Since the electro-optic effect and magneto-optical effect can quickly change the dielectric constant, the MZI formed using materials with electro-optical or magneto-optical effect characteristics can quickly load and refresh its phase parameters, thereby enabling the rapid loading and refreshing of the interaction matrix, thereby further improving the computing efficiency of the optical computing device 10.

[0085] Furthermore, due to the small size of the MZI, the structure of the optical computing device 10 is more compact and can be implemented on a chip, thereby ensuring the stability of the optical computing device 10.

[0086] 3. The first feedback module 300 and the second feedback module 400.

[0087] The first feedback module 300 and the second feedback module 400 may each include a plurality of optical fiber waveguides, each of which may transmit an optical signal.

[0088] The first feedback module 300 is connected to both ends of the parametric oscillator array 100. When the parametric oscillator array 100 outputs the first set of optical signals, the first feedback module 300 feeds back each optical signal in the first set of optical signals to the parametric oscillator array 100. Figure 2 As shown, the number of optical fiber waveguides included in the first feedback module 300 is the same as the number of parametric oscillators 110 included in the parametric oscillator array 100, and each optical fiber waveguide is connected to both ends of a parametric oscillator 110 to feed back the optical signal output by the parametric oscillator 110 to the parametric oscillator 110.

[0089] The second feedback module 400 is connected to the parametric oscillator array 100 and the interaction calculation matrix 200. When the interaction calculation matrix 200 outputs the second group of optical signals, the second feedback module 400 feeds back each optical signal in the second group of optical signals to the parametric oscillator array 100. Figure 2As shown, and the number of optical signals included in the second group of optical signals output by the interaction calculation matrix 200 is the same as the number of parametric oscillators 110 included in the parametric oscillator array 100, the number of optical fiber waveguides included in the second feedback module 400 is the same as the number of parametric oscillators 110 included in the parametric oscillator array 100.

[0090] Alternatively, the second feedback module 400 may further include a fiber combiner or a fiber splitter. For example, if the parametric oscillator array 100 includes N parametric oscillators 110, thereby outputting a first group of optical signals including N optical signals, and after matrix operations of the interaction calculation matrix 200, the second group of optical signals includes 2N optical signals, then a fiber combiner can be provided between any two optical fiber waveguides to combine the optical signals transmitted on the two optical fiber waveguides into a signal in the time domain, thereby obtaining N combined optical signals, which are then fed back to the N parametric oscillators 110.

[0091] Alternatively, the first feedback module 300 and the second feedback module 400 may also include only one optical fiber and a fiber combiner. The fiber combiner may combine the first group of optical signals output by the parametric oscillator array 100 into a signal in the time domain. The combined signal is then fed back to the parametric oscillator array 100 via the optical fiber included in the first feedback module 300. In this case, after the parametric oscillator array 100 receives the combined signal fed back by the first feedback module 300, it may split the combined signal into N signals and then perform processing based on the N signals.

[0092] In the embodiment of the present application, the optical fiber waveguides included in the first feedback module 300 and the second feedback module 400 can be waveguides with low transmission loss characteristics, such as silicon nitride waveguides, or other semiconductor materials with low transmission loss, such as silicon and silicon dioxide. In this way, the loss of the optical signal fed back to the parametric oscillator array 100 through the first feedback module 300 and the second feedback module 400 can be reduced, reducing the distortion of the optical signal during transmission, thereby improving the accuracy of the calculation results of the optical computing device 10.

[0093] In the above embodiment, the optical computing device 10 is described by taking as an example an optical computing device 10 including a parametric oscillator array 100, an interaction computing matrix 200, a first feedback module 300, and a second feedback module 400. However, in other embodiments, the optical computing device 10 may further include other modules. For example, as can be seen from the foregoing, the first group of optical signals (or the third group of optical signals) output by the parametric oscillator array 100 will be fed back to the parametric oscillator array 100 via the first feedback module 300. The first group of optical signals (or the third group of optical signals) will also participate in the matrix operation in the interaction computing matrix 200. Therefore, a splitter may be further provided between the parametric oscillator array 100, the first feedback module 300, and the interaction computing matrix 200.

[0094] Please refer to Figure 8 Taking the example of a parametric oscillator array 100 including one of the multiple parametric oscillators 110, a splitter is provided between the parametric oscillator 110 and the interaction calculation matrix 200, so that a portion of the optical signal output by the parametric oscillator 110 is transmitted to the first feedback module 300, and the other portion is transmitted to the interaction calculation matrix 200. The other parametric oscillators 110 can also be provided with splitters in the same manner, which will not be described in detail here. As an example, the splitter can be a tunable MZI splitter, which uses thermal tuning to control the phase parameters of the MZI, thereby dynamically adjusting the splitting ratio of the MZI splitter. For example, the splitting ratio can be 90 / 10. After passing through the MZI splitter, 90% of the optical signal output by each parametric oscillator 110 is transmitted to the first feedback module 300, and 10% of the optical signal output by each parametric oscillator 110 is transmitted to the interaction calculation matrix 200.

[0095] In addition, since the parametric oscillator array 100 needs to generate the input signal (for example, the third set of optical signals) required for the optical computing device 10 to perform calculations based on the optical signals fed back from the first feedback module 300 and the second feedback module 400, a combiner can also be provided at the input end of the parametric oscillator array 100.

[0096] Please refer to Figure 9 Taking one of the multiple parametric oscillators 110 included in the parametric oscillator array 100 as an example, a beam combiner is provided between the parametric oscillator 110 and the second feedback module 400 to combine the optical signals fed back by the first feedback module 300 and the second feedback module 400 into a single optical signal in the time domain, which is then input into the parametric oscillator 110 for processing. The same method can also be used to provide beam combiners for other parametric oscillators 110, which will not be described in detail here.

[0097] Alternatively, the optical computing device 10 may further include a phase detector, such as a balanced homodyne detector (BHD). The phase detector may be connected to the parametric oscillator array 100 to detect the phase of each optical signal in a group of optical signals output by the parametric oscillator array 100, thereby determining a calculation result of the optical computing device 100 based on the phase of each optical signal. For example, when the detector detects that the phase of each optical signal output by the parametric oscillator array 100 is 0 or π, the final calculation result is determined based on the group of optical signals.

[0098] Of course, the optical computing device 10 may also include other modules, which are not listed one by one here.

[0099] In the optical computing device provided in the embodiments of the present application, since the parametric oscillator array and the interaction computation matrix process multiple optical signals included in a group of optical signals in parallel, even when the number of nodes in the Ising model is large and a large number of input optical signals need to be introduced, the operation time of the optical computing device will not be increased.

[0100] Moreover, by generating optical signals for optical computing devices to perform calculations through parametric oscillator arrays, optical signals with compressed states can be obtained, which can ensure to the greatest extent that the optimal solution obtained by the optical computing device is the global optimal solution, thereby ensuring the calculation accuracy.

[0101] Furthermore, the optical computing device provided by the embodiments of this application has a simple structure and can be implemented on-chip, providing a feasible solution for subsequent clustered optical Ising machines. Furthermore, the entire computing process is implemented via optical signals, which allows for fast signal transmission and significantly improves computing speed. Therefore, the optical computing device provided by the embodiments of this application can be applied to neural network systems, for example, to implement feedback control within these systems.

[0102] The following will be combined with the above embodiments to Figure 1 Taking the optical computing device shown as an example, the optical signal processing method provided in the embodiment of the present application is introduced.

[0103] Please refer to Figure 10 , is a flowchart of a method for processing an optical signal provided in an embodiment of the present application, and the flowchart is described as follows:

[0104] S1001: The parametric oscillator array 100 receives a first group of signals.

[0105] The description of the first group of signals may refer to the above introduction to the parametric oscillator array 100 , which will not be repeated here.

[0106] S1002 : The parametric oscillator array 100 generates a first group of optical signals according to the first group of signals.

[0107] The first group of optical signals includes a plurality of first optical signals. The process of the parametric oscillator array 100 generating the first group of optical signals according to the first group of signals can be referred to the above introduction of the parametric oscillator array 100 and will not be repeated here.

[0108] S1003 : The parametric oscillator array 100 outputs the first group of optical signals, and the interaction calculation array 200 and the first feedback module 300 respectively receive the first group of optical signals.

[0109] The parametric oscillator array 100 may include a plurality of parametric oscillators, the number of which is the same as the number of optical signals included in the first group of optical signals, and each parametric oscillator is configured to output one optical signal in the first group of optical signals.

[0110] As an example, each parametric oscillator receives one optical signal of the first group of optical signals, and divides the optical signal into a first optical signal and a second optical signal according to a preset splitting ratio, and then transmits the first optical signal to the first feedback module 300, and transmits the second optical signal to the interaction calculation matrix 200. For the specific process, please refer to the embodiment of the present invention. Figure 8 The content described.

[0111] S1004 : The interaction calculation array 200 performs a matrix operation on the first group of optical signals according to a preset matrix to obtain a second group of optical signals.

[0112] The second group of optical signals includes a plurality of second optical signals. The specific process of step S1003 can refer to the above introduction of the interaction matrix 200 and will not be repeated here.

[0113] S1005 : The interaction calculation array 200 outputs the second group of optical signals, and the second feedback module 400 receives the second group of optical signals.

[0114] S1006 : The first feedback module 300 transmits the first group of optical signals to the parametric oscillator array 100 .

[0115] S1007 : The second feedback module 400 transmits the second group of optical signals to the parametric oscillator array 100 .

[0116] S1008 : The parametric oscillator array 100 generates a third group of optical signals according to the first group of signals, the first group of optical signals, and the second group of optical signals.

[0117] Step S1008 is similar to step S1002 and will not be described again here.

[0118] It should be noted that each parametric oscillator can combine an optical signal of the first group of optical signals and an optical signal of the second group of optical signals into one optical signal, and transmit the combined optical signal to the parametric oscillator for the above processing. The specific process can be referred to for Figure 9 The content described.

[0119] S1009 : The parametric oscillator array 100 outputs the third group of optical signals, the interaction computing array 200 receives the third group of optical signals, and the first feedback module 300 receives the third group of optical signals.

[0120] Step S1009 is similar to step S1003 and will not be described again here.

[0121] S1010: The interaction calculation array 200 performs a matrix operation on the third group of optical signals according to the preset matrix to obtain a fourth group of optical signals.

[0122] S1011 : The interaction calculation array 200 outputs a fourth group of optical signals, and the second feedback module 400 receives the fourth group of optical signals.

[0123] S1012 : The first feedback module 300 transmits the third group of optical signals to the parametric oscillator array 100 .

[0124] S1013 : The second feedback module 400 transmits the fourth group of optical signals to the parametric oscillator array 100 .

[0125] It is understood that in practical applications, steps S1008 to S1013 may be repeated multiple times, and the number of times each step may be executed is not limited. When the phase of each optical signal in a group of optical signals output by the parametric oscillator array 100 meets a preset condition, for example, the phase of each optical signal is 0 or π, the final calculation result is determined based on the group of optical signals.

[0126] It should be noted that the embodiments provided in this application are merely illustrative. Those skilled in the art will clearly understand that, for the convenience and brevity of description, in the above embodiments, the descriptions of each embodiment have their own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The features disclosed in the embodiments, claims, and drawings of this application may exist independently or in combination. The features described in the embodiment of this application in the form of hardware may be executed by software, and vice versa, which is not limited here.

Claims

1. An optical computing device, characterized in that include: a parametric oscillator array, configured to receive a first group of signals and generate a first group of optical signals according to the first group of signals, wherein the first group of optical signals includes a plurality of first optical signals; an interaction computing array, connected to the parametric oscillator array, configured to receive the first group of optical signals and perform a matrix operation on the first group of optical signals according to a preset matrix to obtain a second group of optical signals, where the second group of optical signals includes a plurality of second optical signals; a first feedback module connected to both ends of the parametric oscillator array, configured to receive the first group of optical signals and transmit the first group of optical signals to the parametric oscillator array; The second feedback module is connected to the parametric oscillator array and the interaction calculation array, and is used to receive the second group of optical signals and transmit the second group of optical signals to the parametric oscillator array.

2. The optical computing device according to claim 1, wherein The parametric oscillator array is also used for: receiving the first group of optical signals and the second group of optical signals; outputting a third group of optical signals according to the first group of signals, the first group of optical signals, and the second group of optical signals; The interaction calculation array is also used to: receiving the third group of optical signals, and performing a matrix operation on the third group of optical signals according to the preset matrix to obtain a fourth group of optical signals; The first feedback module is further configured to: receiving the third group of optical signals, and transmitting the third group of optical signals to the parametric oscillator array; The second feedback module is further configured to: The fourth group of optical signals is received, and the fourth group of optical signals is transmitted to the parametric oscillator array.

3. The optical computing device according to claim 1 or 2, characterized in that The parametric oscillator array includes a plurality of parametric oscillators. The parametric oscillators include a first waveguide and a second waveguide. The first waveguide and the second waveguide are made of different materials. The material of the first waveguide includes a material having a nonlinear effect.

4. The optical computing device according to claim 3, wherein The material of the second waveguide includes a material having a transmission loss lower than a threshold value.

5. The optical computing device according to any one of claims 1, 2, and 4, characterized in that The interaction calculation array includes multiple cascaded Mach-Zehnder (MZ) interferometers, each of which includes multiple Mach-Zehnder interferometers (MZIs) and beam splitters. Each MZI includes a waveguide with a dielectric constant adjustment speed less than a threshold, and the waveguide is used to adjust the phase parameters of the corresponding MZI.

6. The optical computing device according to claim 3, wherein The interaction calculation array includes multiple cascaded Mach-Zehnder (MZ) interferometers, each of which includes multiple Mach-Zehnder interferometers (MZIs) and beam splitters. Each MZI includes a waveguide with a dielectric constant adjustment speed less than a threshold, and the waveguide is used to adjust the phase parameters of the corresponding MZI.

7. The optical computing device according to claim 5, wherein The waveguides in the multiple MZIs included in the same MZ interference unit are made of different materials.

8. The optical computing device according to claim 6, wherein The waveguides in the multiple MZIs included in the same MZ interference unit are made of different materials.

9. The optical computing device according to any one of claims 1, 2, 4, 6, 7, and 8, characterized in that The optical computing device further comprises: An optical splitter is connected to the parametric oscillator, the first feedback module, and the interaction calculation array, and is used to: receiving a first optical signal from the first group of optical signals; Splitting the optical signal into a first optical signal portion and a second optical signal portion according to a preset splitting ratio; The first portion of the optical signal is transmitted to the first feedback module, and the second portion of the optical signal is transmitted to the interaction calculation array.

10. The optical computing device according to claim 3, wherein The optical computing device further comprises: an optical splitter connected to the parametric oscillator, the first feedback module, and the interaction calculation array, and configured to: receiving a first optical signal from the first group of optical signals; Splitting the optical signal into a first optical signal portion and a second optical signal portion according to a preset splitting ratio; The first portion of the optical signal is transmitted to the first feedback module, and the second portion of the optical signal is transmitted to the interaction calculation array.

11. The optical computing device according to claim 5, wherein The optical computing device further comprises: An optical splitter is connected to the parametric oscillator, the first feedback module, and the interaction calculation array, and is used to: receiving a first optical signal from the first group of optical signals; Splitting the optical signal into a first optical signal portion and a second optical signal portion according to a preset splitting ratio; The first portion of the optical signal is transmitted to the first feedback module, and the second portion of the optical signal is transmitted to the interaction calculation array.

12. The optical computing device according to any one of claims 1, 2, 4, 6, 7, 8, 10, and 11, wherein: The optical computing device further comprises: a combiner connected to the parametric oscillator, the first feedback module, and the second feedback module, configured to combine a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmit the combined optical signal to the parametric oscillator.

13. The optical computing device according to claim 3, wherein The optical computing device further comprises: a combiner connected to the parametric oscillator, the first feedback module, and the second feedback module, configured to combine a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmit the combined optical signal to the parametric oscillator.

14. The optical computing device according to claim 5, wherein The optical computing device further comprises: a combiner connected to the parametric oscillator, the first feedback module, and the second feedback module, configured to combine a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmit the combined optical signal to the parametric oscillator.

15. The optical computing device according to claim 9, wherein The optical computing device further comprises: a combiner connected to the parametric oscillator, the first feedback module, and the second feedback module, configured to combine a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmit the combined optical signal to the parametric oscillator.

16. A method for processing an optical signal, characterized in that: The method is performed by an optical computing device, the optical computing device including a parametric oscillator array, an interaction computing array, a first feedback module, and a second feedback module. The method includes: The parametric oscillator array receives a first group of signals, and generates a first group of optical signals according to the first group of signals, wherein the first group of optical signals includes a plurality of first optical signals; The interaction calculation array receives the first group of optical signals and performs a matrix operation on the first group of optical signals according to a preset matrix to obtain a second group of optical signals, where the second group of optical signals includes a plurality of second optical signals; The first feedback module receives the first group of optical signals and transmits the first group of optical signals to the parametric oscillator array; The second feedback module receives the second group of optical signals and transmits the second group of optical signals to the parametric oscillator array.

17. The method according to claim 16, characterized in that The method further comprises: The parametric oscillator array receives the first group of optical signals and the second group of optical signals, and outputs a third group of optical signals according to the first group of signals, the first group of optical signals, and the second group of optical signals; The interaction calculation array receives the third group of optical signals, and performs matrix operations on the third group of optical signals according to the preset matrix to obtain a fourth group of optical signals; The first feedback module receives the third group of optical signals and transmits the third group of optical signals to the parametric oscillator array; The second feedback module receives the fourth group of optical signals and transmits the fourth group of optical signals to the parametric oscillator array.

18. The method according to claim 16 or 17, characterized in that The parametric oscillator array includes a plurality of parametric oscillators, and the method further includes: Each parametric oscillator receives a first optical signal in the first group of optical signals, and splits the first optical signal into a first partial optical signal and a second partial optical signal according to a preset splitting ratio; Each parametric oscillator transmits the first portion of the optical signal to the first feedback module, and transmits the second portion of the optical signal to the interaction calculation array.

19. The method according to claim 16 or 17, characterized in that The parametric oscillator array includes a plurality of parametric oscillators, and the method further includes: Each parametric oscillator combines a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmits the combined optical signal to the parametric oscillator.

20. The method according to claim 18, wherein The parametric oscillator array includes a plurality of parametric oscillators, and the method further includes: Each parametric oscillator combines a first optical signal of the first group of optical signals and a second optical signal of the second group of optical signals into one optical signal, and transmits the combined optical signal to the parametric oscillator.

Images