WO2023045202A1 - 一种光学芯片设计方法、系统、设备及计算机存储介质 - Google Patents

一种光学芯片设计方法、系统、设备及计算机存储介质 Download PDF

Info

Publication number
WO2023045202A1
WO2023045202A1 PCT/CN2022/074612 CN2022074612W WO2023045202A1 WO 2023045202 A1 WO2023045202 A1 WO 2023045202A1 CN 2022074612 W CN2022074612 W CN 2022074612W WO 2023045202 A1 WO2023045202 A1 WO 2023045202A1
Authority
WO
WIPO (PCT)
Prior art keywords
operation function
target
mzi
fitting
optical chip
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2022/074612
Other languages
English (en)
French (fr)
Inventor
吴睿振
王凛
陈静静
黄萍
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou Wave Intelligent Technology Co Ltd
Original Assignee
Suzhou Wave Intelligent Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Suzhou Wave Intelligent Technology Co Ltd filed Critical Suzhou Wave Intelligent Technology Co Ltd
Publication of WO2023045202A1 publication Critical patent/WO2023045202A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/18—Chip packaging

Definitions

  • the present application relates to the field of electronic information technology, and more specifically, relates to an optical chip design method, system, device and computer storage medium.
  • optical computing technology is rapidly increasing.
  • the characteristics of low power consumption and low power consumption are more and more valued by people;
  • the second is the parallel computing characteristics of optical computing technology, as well as the development of algorithms and hardware architectures such as optical neural networks, which provide artificial intelligence technologies such as image recognition, speech recognition, and virtual reality.
  • the need for computing power provides the most potential solution.
  • Optical computing can be divided into analog optical computing and digital optical computing.
  • One of the most typical examples of analog light calculations is Fourier operations, which require the use of Fourier transform-related calculations, such as convolution calculations, in areas such as image processing.
  • Fourier transform-related calculations such as convolution calculations
  • Using a traditional computer to calculate the Fourier transform is very computationally intensive, and the process of light passing through the lens itself is a Fourier transform process, which takes almost no time at all.
  • Digital optical computing uses the combination of light and optical devices to form classic logic gates, constructs a computing system similar to traditional digital electronic computing principles, and realizes computing through complex logic gate combination operations.
  • optical chips generated by optical computing can process data quickly, however, existing optical chips still need to convert optical signals into electrical signals in the process of realizing nonlinear operations, resulting in complex structures of existing optical chips and low operational efficiency. Low.
  • the purpose of the present application is to provide an optical chip design method, which can solve the technical problem of how to improve the operational efficiency of the optical chip to achieve nonlinear operations to a certain extent.
  • the present application also provides an optical chip design system, equipment and computer-readable storage medium.
  • a method for designing an optical chip comprising:
  • the target optical chip is based on the MZI design
  • the structure of the target optical chip includes the number of the MZI, connection relationship and configuration parameters, so as to base on the target The structure of the optical chip designs the target optical chip.
  • determining the fitting connection relationship of the MZI based on the independent variable in the target nonlinear operation function and the structure of the MZI includes:
  • a connection relationship of the output whose characteristics are consistent with the target nonlinear operation function is determined as the fitting connection relationship.
  • the determining the fitting linear operation function of the fitting nonlinear operation function based on the least squares method includes:
  • the fitting linear operation function corresponding to the fitting nonlinear operation function is determined based on the design accuracy and the least square method.
  • the design precision includes the data set range of the independent variable
  • the determining the fitting linear operation function corresponding to the fitting nonlinear operation function based on the design accuracy and the least squares method includes:
  • the initial linear operation function with the largest similarity value is used as the fitted linear operation function.
  • determining the structure of the target optical chip based on the fitted linear operation function and the fitted connection relationship includes:
  • the structure of the target optical chip is determined based on the fitting linear operation function, the fitting connection relationship and a singular value decomposition method.
  • the determining the structure of the target optical chip based on the fitting linear operation function, the fitting connection relationship and the singular value decomposition method includes:
  • the structure of the target optical chip is determined based on the matrix decomposition result.
  • the target nonlinear operation function includes a sigmoid function
  • the structure of the target optical chip includes: a first MZI, a photodiode, a first amplifier, a phase shift controller, a second MZI, and a target topology; the first optical input end and the second optical input end of the first MZI connected, the second optical output end of the first MZI is suspended, the first optical output end of the first MZI is connected to the input end of the photodiode, the first optical input end of the second MZI and the target
  • the second optical input ends of the topology are respectively connected; the output ends of the photodiodes are connected to the first amplifier; the first amplifier is connected to the phase shift controller; the phase shift controller is connected to the first The phase shifters of the two MZIs are connected; the second optical input end and the second optical output end of the second MZI are suspended; the first optical output end of the second MZI is connected to the first optical input end of the target topology end connection.
  • An optical chip design system comprising:
  • the first acquisition module is used to acquire the target nonlinear operation function to be realized by the target optical chip, the target optical chip is based on the MZI design;
  • the first determination module is used to determine the fitting connection relationship of the MZI based on the independent variable in the target nonlinear operation function and the structure of the MZI;
  • a first conversion module configured to convert the target nonlinear operation function into a fitted nonlinear operation function to be fitted by the MZI based on the fitting connection relationship;
  • the second determination module is used to determine the fitting linear operation function corresponding to the fitting nonlinear operation function based on the least square method
  • a third determining module configured to determine the structure of the target optical chip based on the fitted linear operation function and the fitted connection relationship, where the structure of the target optical chip includes the number, connection relationship and configuration of the MZIs parameters to design the target optical chip based on the structure of the target optical chip.
  • the determining the fitting linear operation function of the fitting nonlinear operation function based on the least squares method includes:
  • the fitting linear operation function corresponding to the fitting nonlinear operation function is determined based on the design accuracy and the least square method.
  • the design precision includes the data set range of the independent variable
  • the determining the fitting linear operation function corresponding to the fitting nonlinear operation function based on the design accuracy and the least squares method includes:
  • the initial linear operation function with the largest similarity value is used as the fitted linear operation function.
  • An optical chip design device comprising:
  • a processor configured to implement the steps of any one of the above optical chip design methods when executing the computer program.
  • a computer-readable storage medium wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the steps of any one of the above-mentioned optical chip design methods are realized.
  • a method for designing an optical chip obtaineds the target nonlinear operation function to be realized by the target optical chip, and the target optical chip is designed based on MZI; based on the independent variable in the target nonlinear operation function and the structure of MZI, the pseudo-linear operation of the MZI is determined.
  • the structure of the target optical chip includes the number of MZIs, the connection relationship and configuration parameters, so as to design the target optical chip based on the structure of the target optical chip.
  • the fitting connection relationship of the MZI can be determined based on the independent variables in the target nonlinear operation function and the structure of the MZI, so that the fitting connection relationship of the MZI is compatible with the target nonlinear operation function; it can be based on the fitting connection relationship , convert the target nonlinear operation function into the fitting nonlinear operation function to be fitted by MZI, so that the fitting nonlinear operation function to be realized by MZI adapts to its own structure; and the fitting nonlinear operation can be determined based on the least square method
  • the fitting linear operation function corresponding to the function, based on the fitting linear operation function and the fitting connection relationship determines the structure of the target optical chip, so that the operation characteristics of the target optical chip are compatible with the connection structure of the MZI, ensuring that the target optical chip is based on the MZI
  • Fig. 1 is the flow chart of a kind of optical chip design method provided by the embodiment of the present application.
  • FIG. 2 is a schematic structural diagram of the MZI
  • Figure 3 is a schematic diagram of the direct connection structure of the MZI
  • Fig. 4 is the comparison structure diagram of each fitting result and sigmoid function
  • Fig. 5 is the partial structure schematic diagram of the optical chip that realizes sigmoid function
  • Fig. 6 is the structural representation of the optical chip realizing sigmoid function
  • FIG. 7 is a schematic structural diagram of an optical chip design system provided by an embodiment of the present application.
  • FIG. 8 is a schematic structural diagram of an optical chip design device provided in an embodiment of the present application.
  • FIG. 9 is another schematic structural diagram of an optical chip design device provided in an embodiment of the present application.
  • FIG. 10 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present application.
  • FIG. 1 is a flow chart of an optical chip design method provided in an embodiment of the present application.
  • Step S101 Obtain the target nonlinear operation function to be realized by the target optical chip, the target optical chip is designed based on Mach-Zehnder interferometer.
  • the target optical chip is used to realize the corresponding target operation function, the application needs to obtain the target nonlinear operation function to be realized by the target optical chip first, and the type of the target nonlinear operation function can be determined according to actual needs.
  • the target nonlinear operation function can be a sigmoid function, etc.
  • the target optical chip in this application can be designed based on MZI (Mach–Zehnder interferometer, Mach-Zehnder interferometer), and the Mach-Zehnder interferometer described here is a An interferometer can be used to observe the relative phase shift changes produced by different paths and media after the beam emitted from a single light source is split into two collimated beams.
  • Step S102 Based on the independent variables in the target nonlinear operation function and the structure of the Mach-Zehnder interferometer, determine the fitting connection relationship of the Mach-Zehnder interferometer.
  • the MZI has two inputs and two outputs, and the connection relationship between the two outputs is different, after obtaining the target nonlinear operation function, it is necessary to base on the independent variables in the target nonlinear operation function and the structure of the MZI, Determine the fitted connectivity relationship for the MZI.
  • the structure of the MZI can be shown in Figure 2, where L 1 and L 2 represent the two optical input ends of the MZI, 2 ⁇ represents the phase shifter of the MZI, and L′ 1 and L′ 2 represent the two optical inputs of the MZI.
  • the optical output terminal; the calculation relationship between the optical output terminal and the optical input terminal of the MZI can be as follows:
  • the derivation process is as follows: Assume that the amplitude of the light input to the MZI is A, the frequency is ⁇ , the time is t, and the initial phases of the two optical input ports are ⁇ 1 and ⁇ 2 respectively, then the expressions of the two optical input ports of the MZI can be:
  • Re represents the real part
  • j represents the imaginary unit
  • the following methods can be used to realize the programmable phase shifter, such as plating a metal film on a section of waveguide material, and controlling the metal film heater by applying an external voltage to cause the waveguide Temperature changes to change the refractive index to achieve phase shift; for example, using the plasma diffusion effect (changing the concentration of electrons and holes) and electro-optic effect to change the waveguide refractive index to introduce phase shift, etc.; this application does not make specific limitations here.
  • Step S103 Based on the fitting connection relationship, convert the target nonlinear operation function into a fitting nonlinear operation function to be fitted by the Mach-Zehnder interferometer.
  • the target nonlinear operation function is not compatible with the structure of MZI
  • after determining the fitting connection relationship of MZI based on the independent variables in the target nonlinear operation function and the structure of MZI it is also necessary to base on the fitting
  • the connection relationship converts the target nonlinear operation function into the fitting nonlinear operation function to be fitted by MZI.
  • Step S104 Determine the fitting linear operation function corresponding to the fitting nonlinear operation function based on the least square method.
  • MZI can only achieve linear multiply-add operations, in the process of realizing the target nonlinear operation function based on the target optical chip, it is also necessary to determine the fitted linear operation function corresponding to the fitted nonlinear operation function based on the least square method , in addition, because the fitting nonlinear operation function can make the target nonlinear operation function adapt to the structure of MZI, after determining the fitting linear operation function corresponding to the fitting nonlinear operation function based on the least squares method, it can ensure that the MZI achieves the target Non-linear operation function.
  • Step S105 Determine the structure of the target optical chip based on the fitting linear operation function and the fitting connection relationship. Design the target optical chip.
  • the fitted connection relationship only shows that the MZI achieves the goal
  • the structure used by the nonlinear operation function, the configuration parameters in the MZI are uncertain, so it is still necessary to determine the target optical chip based on the fitted linear operation function and the fitted connection relationship.
  • the structure of the target optical chip may include the number of Mach-Zehnder interferometers, connection relations and configuration parameters, so that based on the structure of the target optical chip, the target optical chip that can realize the target nonlinear operation function is designed.
  • a method for designing an optical chip obtaineds the target nonlinear operation function to be realized by the target optical chip, and the target optical chip is designed based on MZI; based on the independent variable in the target nonlinear operation function and the structure of MZI, the pseudo-linear operation of the MZI is determined.
  • the structure of the target optical chip includes the number of MZIs, the connection relationship and configuration parameters, so as to design the target optical chip based on the structure of the target optical chip.
  • the fitting connection relationship of the MZI can be determined based on the independent variables in the target nonlinear operation function and the structure of the MZI, so that the fitting connection relationship of the MZI is compatible with the target nonlinear operation function; it can be based on the fitting connection relationship , convert the target nonlinear operation function into the fitting nonlinear operation function to be fitted by MZI, so that the fitting nonlinear operation function to be realized by MZI adapts to its own structure; and the fitting nonlinear operation can be determined based on the least square method
  • the fitting linear operation function corresponding to the function, based on the fitting linear operation function and the fitting connection relationship determines the structure of the target optical chip, so that the operation characteristics of the target optical chip are compatible with the connection structure of the MZI, ensuring that the target optical chip is based on the MZI
  • this process may include the following steps: use the independent variable in the target nonlinear operation function as the input of MZI, determine the connection relationship between the two outputs of MZI based on the structure of MZI; connect the output of the feature consistent with the target nonlinear operation function , determined as the fitted connection relationship.
  • the target nonlinear operation function is still the sigmoid function as an example, and its expression is: Because there is only one argument at this time, one input port of MZI needs to be sealed. At this time, the structure of MZI is shown in Figure 3, and the corresponding output after the argument of the sigmoid function is input into MZI is: Because the only thing that can be observed through the MZI is no in the output, the corresponding output after the independent variable is input into the MZI can be equivalent to Finally, it can be seen from the curve characteristics of the sigmoid function that the function value cannot be less than 0, and only the connection relationship -xsin ⁇ cos ⁇ in MZI can be fitted, so the fitting connection relationship of the sigmoid function in MZI is: At this time, in the process of converting the target nonlinear operation function into the fitted nonlinear operation function to be fitted by MZI based on the fitting connection relationship, because -xsin ⁇ cos ⁇ is equivalent to xsin2 ⁇ , this function
  • the fitting linear operation function can be determined according to the accuracy, that is, the design accuracy of the target optical chip can be obtained; based on the design accuracy and The least squares method determines the fitting linear operation function corresponding to the fitting nonlinear operation function.
  • the design accuracy can be characterized by the range of the independent variable, that is, the design accuracy can include the data set range of the independent variable;
  • the initial linear operation of the fitting nonlinear operation function at each power can be determined based on the least square method function; within the scope of the data set, determine the similarity between each initial linear operation function and the fitted nonlinear operation function; use the initial linear operation function with the largest similarity value as the fitted linear operation function.
  • the fitting result can be:
  • the structure of the target optical chip in the process of determining the structure of the target optical chip based on the fitting linear operation function and fitting connection relationship, because in the application process of MZI, it can be used in the optical domain Singular Value Decomposition (SVD) is used to realize matrix multiplication of any dimension, so in this application, the structure of the target optical chip can be determined based on the fitting linear operation function, fitting connection relationship and singular value decomposition method.
  • Singular Value Decomposition singular value Decomposition
  • the structure of the target optical chip can be determined based on the fitting linear operation function and the fitting connection relationship.
  • a link arithmetic matrix decompose the link arithmetic matrix based on a singular value decomposition method to obtain a matrix decomposition result; determine the structure of the target optical chip based on the matrix decomposition result.
  • the phase modulation is controlled, so that an adjustment relationship from
  • 2 2 ⁇ ; further, the output in Figure 5
  • 2 0.0018x 3
  • 2 -0.2533x-1; because -0.2533 in the final mapping relationship of the target optical chip x can be easily realized in the MZI structure in Figure 3 by configuring ⁇ (arcsin(-0.2533))/2, and -1 in the final mapping relationship of the target optical chip can be decomposed by SVD through MZI by introducing fixed optical information
  • the structure of the target optical chip can be as shown in Figure 6, which can include: a first MZI11, a photodiode 12, a first amplifier 13, a phase shift controller 14, and a second MZI15 , the target topology 16; the first optical input end of the first MZI is connected to the second optical input end, the second optical output end of the first MZI is suspended, the first optical output end of the first MZI is connected to the input end of the photodiode, The first optical input end of the second MZI is connected to the second optical input end of the target topology respectively; the output end of the photodiode is connected to the first amplifier; the first amplifier is connected to the phase-shift controller; the phase-shift controller is connected to the second The phase shifter of the MZI is connected; the second optical input end and the second optical output end of the second MZI are suspended; the first optical output end of the second MZI is connected to the first optical input end of the
  • the target topology may include the third MZI161, the fourth MZI162, the fifth MZI163, and the laser transmitter.
  • the first light of the target topology The input end is the first optical input end of the third MZI
  • the second optical input end of the target topology is the second optical input end of the third MZI
  • the optical output end of the target topology is the first optical output end of the fifth MZI
  • the first optical output end of the third MZI is connected to the first optical input end of the fifth MZI
  • the second optical output end of the third MZI is connected to the first optical input end of the fourth MZI
  • the laser transmitter is connected to the fourth MZI connected to the second optical input end of the fourth MZI
  • the first optical output end of the fourth MZI is connected to the second optical input end of the fifth MZI
  • the second optical output end of the fourth MZI is suspended
  • the second optical output end of the fifth MZI is suspended in the air .
  • FIG. 7 is a schematic structural diagram of an optical chip design system provided by an embodiment of the present application.
  • the first obtaining module 101 is used to obtain the target nonlinear operation function to be realized by the target optical chip, and the target optical chip is designed based on MZI;
  • the first determination module 102 is used to determine the fitting connection relationship of the MZI based on the independent variables in the target nonlinear operation function and the structure of the MZI;
  • the first conversion module 103 is used to convert the target nonlinear operation function into a fitted nonlinear operation function to be fitted by the MZI based on the fitting connection relationship;
  • the second determination module 104 is used to determine the fitting linear operation function corresponding to the fitting nonlinear operation function based on the least squares method
  • the third determination module 105 is used to determine the structure of the target optical chip based on the fitting linear operation function and the fitting connection relationship. Design the target optical chip.
  • the first determination module may include:
  • the first determination unit is used to use the independent variable in the target nonlinear operation function as the input of the MZI, and determine the connection relationship between the two outputs of the MZI based on the structure of the MZI;
  • the second determining unit is configured to determine the connection relationship of the output whose features are consistent with the target nonlinear operation function as a fitting connection relationship.
  • the second determination module may include:
  • the first acquisition unit is used to acquire the design accuracy of the target optical chip
  • the third determination unit is configured to determine the fitting linear operation function corresponding to the fitting nonlinear operation function based on the design accuracy and the least square method.
  • the design accuracy includes the data set range of the independent variable
  • the third determination unit can be specifically used to: determine the initial linear operation function of the fitted nonlinear operation function at each power based on the least square method; determine each initial linear operation function and the fitted nonlinear operation function within the scope of the data set The similarity between them; the initial linear operation function with the largest similarity value is used as the fitting linear operation function.
  • the third determination module may include:
  • the third determination unit is used to determine the structure of the target optical chip based on the fitting linear operation function, fitting connection relationship and singular value decomposition method.
  • the third determining unit can be specifically used to: determine the link arithmetic matrix of the target optical chip based on the fitting linear operation function and the fitting connection relationship; The arithmetic matrix is decomposed to obtain a matrix decomposition result; the structure of the target optical chip is determined based on the matrix decomposition result.
  • the target nonlinear operation function includes a sigmoid function
  • the structure of the target optical chip includes: a first MZI, a photodiode, a first amplifier, a phase shift controller, a second MZI, and a target Topological structure; the first optical input end of the first MZI is connected to the second optical input end, the second optical output end of the first MZI is suspended, the first optical output end of the first MZI is connected to the input end of the photodiode, and the second MZI
  • the first optical input end of the target topology is connected to the second optical input end of the target topology; the output end of the photodiode is connected to the first amplifier; the first amplifier is connected to the phase shift controller; the phase shift controller is connected to the second MZI's shift The phase controllers are connected; the second optical input end and the second optical output end of the second MZI are suspended; the first optical output end of the second MZI is connected to the
  • the types of the target topology include: Gridnet topology and Reck topology.
  • FIG. 8 is a schematic structural diagram of an optical chip design device provided in an embodiment of the present application.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202, a computer program is stored in the memory 201, and the processor 202 implements the following steps when executing the computer program:
  • the target nonlinear operation function is converted into a fitting nonlinear operation function to be fitted by the MZI;
  • the structure of the target optical chip is determined.
  • the structure of the target optical chip includes the number of MZIs, the connection relationship and configuration parameters, so as to design the target optical chip based on the structure of the target optical chip.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202.
  • a computer program is stored in the memory 201.
  • the processor 202 executes the computer program, the following steps are implemented: As the input of the MZI, the connection relationship between the two outputs of the MZI is determined based on the structure of the MZI; the connection relationship of the output whose characteristics are consistent with the target nonlinear operation function is determined as the fitting connection relationship.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202.
  • a computer program is stored in the memory 201.
  • the processor 202 executes the computer program, the following steps are implemented: obtaining the design accuracy of the target optical chip; The accuracy and the least squares method determine the fitting linear operation function corresponding to the fitting nonlinear operation function.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202.
  • a computer program is stored in the memory 201.
  • the processor 202 executes the computer program, the following steps are implemented: the design precision includes the data set range of the independent variable; Determine the initial linear operation function of the fitted nonlinear operation function at each power based on the least square method; within the scope of the data set, determine the similarity between each initial linear operation function and the fitted nonlinear operation function; the similarity value The largest initial linear operation function is used as the fitted linear operation function.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202.
  • a computer program is stored in the memory 201.
  • the processor 202 executes the computer program, the following steps are implemented: based on a fitting linear operation function, a fitting connection Relation and singular value decomposition method to determine the structure of the target optical chip.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202.
  • a computer program is stored in the memory 201.
  • the processor 202 executes the computer program, the following steps are implemented: based on the fitting linear operation function and the fitting connection
  • the link arithmetic matrix of the target optical chip is determined; the link arithmetic matrix is decomposed based on the singular value decomposition method to obtain a matrix decomposition result; the structure of the target optical chip is determined based on the matrix decomposition result.
  • An optical chip design device provided in an embodiment of the present application includes a memory 201 and a processor 202.
  • a computer program is stored in the memory 201.
  • the processor 202 executes the computer program, the following steps are implemented: the target nonlinear operation function includes a sigmoid function; the target
  • the structure of the optical chip includes: a first MZI, a photodiode, a first amplifier, a phase shift controller, a second MZI, and a target topology; the first optical input end of the first MZI is connected to the second optical input end, and the first MZI The second optical output end of the first MZI is suspended, and the first optical output end of the first MZI is connected to the input end of the photodiode, the first optical input end of the second MZI, and the second optical input end of the target topology; the output of the photodiode
  • the terminal is connected to the first amplifier; the first amplifier is connected to the phase shift controller; the phase shift controller is connected to the
  • another optical chip design device may also include: an input port 203 connected to the processor 202 for transmitting externally input commands to the processor 202; connected to the processor 202
  • the display unit 204 is used to display the processing results of the processor 202 to the outside world;
  • the communication module 205 connected to the processor 202 is used to realize the communication between the optical chip design equipment and the outside world.
  • the display unit 204 can be a display panel, a laser scanning display, etc.; the communication methods adopted by the communication module 205 include but are not limited to mobile high-definition link technology (HML), universal serial bus (USB), high-definition multimedia interface (HDMI), Wireless connection: wireless fidelity technology (WiFi), bluetooth communication technology, low power consumption bluetooth communication technology, communication technology based on IEEE802.11s.
  • HML mobile high-definition link technology
  • USB universal serial bus
  • HDMI high-definition multimedia interface
  • WiFi wireless fidelity technology
  • WiFi wireless fidelity technology
  • Low power consumption bluetooth communication technology communication technology based on IEEE802.11s.
  • FIG. 10 is a schematic structural diagram of a computer-readable storage medium provided in an embodiment of the present application.
  • the computer-readable storage medium 601 stores a computer Program 610, when the computer program 610 is executed by the processor, the following steps are implemented:
  • the target nonlinear operation function is converted into a fitting nonlinear operation function to be fitted by the MZI;
  • the structure of the target optical chip is determined.
  • the structure of the target optical chip includes the number of MZIs, the connection relationship and configuration parameters, so as to design the target optical chip based on the structure of the target optical chip.
  • a computer-readable storage medium provided by an embodiment of the present application, in which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: taking the argument in the target nonlinear operation function as the input of the MZI , determine the connection relationship between the two outputs of the MZI based on the structure of the MZI; determine the connection relationship of the output whose characteristics are consistent with the target nonlinear operation function as the fitting connection relationship.
  • a computer-readable storage medium provided by an embodiment of the present application, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the following steps are implemented: obtaining the design accuracy of the target optical chip; The multiplication determines the fitted linear arithmetic function corresponding to the fitted nonlinear arithmetic function.
  • a computer-readable storage medium provided in an embodiment of the present application, in which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: the design accuracy includes the range of the data set of the independent variable; based on the least squares method Determine the initial linear operation function of the fitted nonlinear operation function at each power; within the scope of the data set, determine the similarity between each initial linear operation function and the fitted nonlinear operation function; the initial linear operation function with the largest similarity value The operation function is used as a fitting linear operation function.
  • a computer-readable storage medium provided by an embodiment of the present application, in which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: based on fitting linear operation functions, fitting connection relations and singular values Decomposition method to determine the structure of the target optical chip.
  • a computer-readable storage medium provided in an embodiment of the present application, in which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: determine the target based on the fitting linear operation function and the fitting connection relationship The link arithmetic matrix of the optical chip; the link arithmetic matrix is decomposed based on the singular value decomposition method to obtain the matrix decomposition result; the structure of the target optical chip is determined based on the matrix decomposition result.
  • the target nonlinear operation function includes a sigmoid function
  • the structure of the target optical chip Including: the first MZI, photodiode, first amplifier, phase shift controller, second MZI, target topology; the first optical input end of the first MZI is connected to the second optical input end, the second optical input end of the first MZI The output end is suspended, and the first optical output end of the first MZI is connected to the input end of the photodiode, the first optical input end of the second MZI and the second optical input end of the target topology; the output end of the photodiode is connected to the first optical input end of the photodiode.
  • the amplifier is connected; the first amplifier is connected with the phase shift controller; the phase shift controller is connected with the phase shifter of the second MZI; the second optical input end and the second optical output end of the second MZI are suspended; the second MZI first An optical output terminal is connected to the first optical input terminal of the target topology; wherein, the types of the target topology include: Gridnet topology and Reck topology.
  • the computer-readable storage medium involved in this application includes random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM , or any other form of storage medium known in the technical field.
  • RAM random access memory
  • ROM read-only memory
  • electrically programmable ROM electrically erasable programmable ROM
  • registers hard disk, removable disk, CD-ROM , or any other form of storage medium known in the technical field.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Optical Modulation, Optical Deflection, Nonlinear Optics, Optical Demodulation, Optical Logic Elements (AREA)

Abstract

本申请公开了一种光学芯片设计方法、系统、设备及计算机可读存储介质，获取目标非线性运算函数（S101）；基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系（S102）；基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数（S103）；基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数（S104）；基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括MZI的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片（S105）。本申请中保证了目标光学芯片基于MZI实现目标非线性运算函数的准确性，且整个光学芯片设计过程中无需将光信号转换为电信号，结构简便，运算效率高。

Description

一种光学芯片设计方法、系统、设备及计算机存储介质
本申请要求在2021年09月23日提交中国专利局、申请号为202111110483.7、发明名称为“一种光学芯片设计方法、系统、设备及计算机存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及电子信息技术领域，更具体地说，涉及一种光学芯片设计方法、系统、设备及计算机存储介质。
背景技术
随着通信技术发展，人们对光计算技术的需求迅速增加，其原因在于：一是随着摩尔定律的逐步失效，以及大数据时代对计算系统功耗和速度要求的不断提高，光计算技术高速和低功耗的特点越来越受到人们的重视；二是光计算技术的并行性运算特点，以及光学神经网络等算法和硬件架构的发展，为图像识别、语音识别、虚拟现实等人工智能技术对算力的需求提供了最有潜力的解决方案。光计算可以分为模拟光计算和数字光计算。模拟光计算最典型的一个例子就是傅立叶运算，在图像处理等领域需要运用傅立叶变换相关的计算，如卷积计算。用传统计算机来计算傅立叶变换是非常消耗计算量的，而光通过透镜的过程，本身就是一次傅立叶变换的过程，这个过程几乎完全不需要时间。数字光计算是利用光和光学器件组合形成经典的逻辑门，构建类似传统数字电子计算原理的计算系统，通过复杂的逻辑门组合操作实现计算。
虽然通过光计算生成的光学芯片可以快速对数据进行处理，然而，现有光学芯片在实现非线性运算的过程中还需要将光信号转换为电信号，导致现有光学芯片的结构复杂，运算效率低。
综上所述，如何提高光学芯片实现非线性运算的运算效率是目前本领域技术人员亟待解决的问题。
发明内容
本申请的目的是提供一种光学芯片设计方法，其能在一定程度上解决如何提高光学芯片实现非线性运算的运算效率的技术问题。本申请还提供了一种光学芯片设计系统、设备及计算机可读存储介质。
为了实现上述目的，本申请提供如下技术方案：
一种光学芯片设计方法，包括：
获取目标光学芯片待实现的目标非线性运算函数，所述目标光学芯片基于MZI设计；
基于所述目标非线性运算函数中的自变量及所述MZI的结构，确定所述MZI的拟合连接关系；
基于所述拟合连接关系，将所述目标非线性运算函数转换为所述MZI待拟合的拟合非线性运算函数；
基于最小二乘法确定所述拟合非线性运算函数对应的拟合线性运算函数；
基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的结构，所述目标光学芯片的结构包括所述MZI的数量、连接关系及配置参数，以基于所述目标光学芯片的结构设计所述目标光学芯片。
可选的，所述基于所述目标非线性运算函数中的自变量及所述MZI的结构，确定所述MZI的拟合连接关系，包括：
将所述目标非线性运算函数中的自变量作为所述MZI的输入，基于所述MZI的结构确定所述MZI两个输出的连接关系；
将特征与所述目标非线性运算函数一致的所述输出的连接关系，确定为所述拟合连接关系。
可选的，所述基于最小二乘法确定所述拟合非线性运算函数的拟合线性运算函数，包括：
获取所述目标光学芯片的设计精度；
基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数。
可选的，所述设计精度包括所述自变量的数据集范围；
所述基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数，包括：
基于所述最小二乘法确定所述拟合非线性运算函数在各个次方时的初始线性运算函数；
在所述数据集范围内，确定各个所述初始线性运算函数与所述拟合非线性运算函数间的相似度；
将相似度值最大的所述初始线性运算函数作为所述拟合线性运算函数。
可选的，所述基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的结构，包括：
基于所述拟合线性运算函数、所述拟合连接关系及奇异值分解法，确定所述目标光学芯片的结构。
可选的，所述基于所述拟合线性运算函数、所述拟合连接关系及奇异值分解法，确定所述目标光学芯片的结构，包括：
基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的链接算术矩阵；
基于所述奇异值分解法对所述链接算术矩阵进行分解，得到矩阵分解结果；
基于所述矩阵分解结果确定所述目标光学芯片的结构。
可选的，所述目标非线性运算函数包括sigmoid函数；
所述目标光学芯片的结构包括：第一MZI、光电二极管、第一放大器、移相控制器、第二MZI、目标拓扑结构；所述第一MZI的第一光输入端与第二光输入端相连，所述第一MZI的第二光输出端悬空，所述第一MZI的第一光输出端与所述光电二极管的输入端、所述第二MZI的第一光输入端及所述目标拓扑结构的第二光输入端分别相连；所述光电二极管的输出端与所述第一放大器相连；所述第一放大器与所述移相控制器相连；所述移相控制器与所述第二MZI的移相器相连；所述第二MZI的第二光输入端及第二光输出端均悬空；所述第二MZI的第一光输出端与所述目标拓扑结构的第一光输入端连接。
一种光学芯片设计系统，包括：
第一获取模块，用于获取目标光学芯片待实现的目标非线性运算函数，所述目标光学芯片基于MZI设计；
第一确定模块，用于基于所述目标非线性运算函数中的自变量及所述MZI的结构，确定所述MZI的拟合连接关系；
第一转换模块，用于基于所述拟合连接关系，将所述目标非线性运算函数转换为所述MZI待拟合的拟合非线性运算函数；
第二确定模块，用于基于最小二乘法确定所述拟合非线性运算函数对应的拟合线性运算函数；
第三确定模块，用于基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的结构，所述目标光学芯片的结构包括所述MZI的数量、连接关系及配置参数，以基于所述目标光学芯片的结构设计所述目标光学芯片。
可选的，所述基于最小二乘法确定所述拟合非线性运算函数的拟合线性运算函数，包括：
获取所述目标光学芯片的设计精度；
基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数。
可选的，所述设计精度包括所述自变量的数据集范围；
所述基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数，包括：
基于所述最小二乘法确定所述拟合非线性运算函数在各个次方时的初始线性运算函数；
在所述数据集范围内，确定各个所述初始线性运算函数与所述拟合非线性运算函数间的相似度；
将相似度值最大的所述初始线性运算函数作为所述拟合线性运算函数。
一种光学芯片设计设备，包括：
存储器，用于存储计算机程序；
处理器，用于执行所述计算机程序时实现如上任一所述光学芯片设计方法的步骤。
一种计算机可读存储介质，所述计算机可读存储介质中存储有计算机程序，所述计算机程序被处理器执行时实现如上任一所述光学芯片设计方法的步骤。
本申请提供的一种光学芯片设计方法，获取目标光学芯片待实现的目标非线性运算函数，目标光学芯片基于MZI设计；基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系；基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数；基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数；基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括MZI的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片。本申请中，可以基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系，使得MZI的拟合连接关系与目标非线性运算函数相适应；可以基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数，使得MZI待实现的拟合非线性运算函数与自身结构相适应；并且可以基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数，基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，使得目标光学芯片的运算特性与MZI的连接架构相适应，保证了目标光学芯片基于MZI实现目标非线性运算函数的准确性，且整个光学芯片设计过程中无需将光信号转换为电信号，结构简便，运算效率高。本申请提供的一种光学芯片设计系统、设备及计算机可读存储介质也解决了相应技术问题。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据提供的附图获得其他的附图。
图1为本申请实施例提供的一种光学芯片设计方法的流程图；
图2为MZI的结构示意图；
图3为MZI的直连结构示意图；
图4为各个拟合结果与sigmoid函数的比较结构图；
图5为实现sigmoid函数的光学芯片的部分结构示意图；
图6为实现sigmoid函数的光学芯片的结构示意图；
图7为本申请实施例提供的一种光线芯片设计系统的结构示意图；
图8为本申请实施例提供的一种光学芯片设计设备的结构示意图；
图9为本申请实施例提供的一种光学芯片设计设备的另一结构示意图；
图10为本申请实施例提供的一种计算机可读存储介质的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
请参阅图1，图1为本申请实施例提供的一种光学芯片设计方法的流程图。
本申请实施例提供的一种光学芯片设计方法，可以包括以下步骤：
步骤S101：获取目标光学芯片待实现的目标非线性运算函数，目标光学芯片基于马赫-曾德尔干涉仪设计。
实际应用中，因为目标光学芯片是用于实现相应的目标运算函数的，所以本申请需先获取目标光学芯片待实现的目标非线性运算函数，目标非线性运算函数的类型可以根据实际需要确定，比如目标非线性运算函数可以为sigmoid函数等，且本申请中的目标光学芯片可以基于MZI(Mach–Zehnder interferometer，马赫-曾德尔干涉仪)设计，这里所描述的马赫-曾德尔干涉仪是一种干涉仪，可以用来观测从单独光源发射的光束分裂成两道准直光束之后，经过不同路径与介质所产生的相对相移变化。
步骤S102：基于目标非线性运算函数中的自变量及马赫-曾德尔干涉仪的结构，确定马赫-曾德尔干涉仪的拟合连接关系。
实际应用中，因为MZI有两个输入及两个输出，且两个输出的连接关系不同，所以在获取目标非线性运算函数之后，需基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系。
需要说明的是，MZI的结构可以如图2所示，其中L 1、L 2表示MZI的两个光输入端，2θ表示MZI的移相器，L′ 1、L′ 2表示MZI的两个光输出端；MZI的光输出端与光输入端间的计算关系可以如下：
Figure PCTCN2022074612-appb-000001
其推导过程如下：假设输入MZI的光的振幅为A，频率为ω，时间为t，两个光输入端的初始相位分别为θ 1和θ 2，则MZI两个光输入端的表达式可以为：
Figure PCTCN2022074612-appb-000002
因为在光电转换中，可以识别的仅有光的实部，虚部表示的是传输过程中的能量损耗，因此上式可以进一步通过欧拉公式转换为：
Figure PCTCN2022074612-appb-000003
其中Re表示实部，j表示虚数单位，
Figure PCTCN2022074612-appb-000004
表示L 1的复杂状态表达式，
Figure PCTCN2022074612-appb-000005
表示L 2的复杂状态表达式；
因为L 1和L 2在进入MZI后，光中所含能量会通过耦合器转移到输出的两个端口对应的光信号上，因此其振幅则为原本的
Figure PCTCN2022074612-appb-000006
因此MZI中的S1、S2与L 1、L 2间的关系可以通过公式表示为：
Figure PCTCN2022074612-appb-000007
基于耦合器关系，可知
Figure PCTCN2022074612-appb-000008
与[S1,S2] T的对应关系为：
Figure PCTCN2022074612-appb-000009
再基于MZI的衰减器工作方式以及MZI的结构，可以得到S3、S4的表达式为：
Figure PCTCN2022074612-appb-000010
结合以上表达式可知MZI的光输出端与光输入端间的计算关系可以如下：
Figure PCTCN2022074612-appb-000011
此外，在需要对MZI的移相器进行控制的时候，可以使用以下方式来实现可编程的移相器，比如在一段波导材料上镀上金属薄膜，通过施加外部电压控制金属薄膜加热器导致波导温度变化来改变折射率，实现相移；比如利用等离子体弥散效应(改变电子和空穴的浓度)和电光效应改变波导折射率来引入相移等；本申请在此不做具体限定。
步骤S103：基于拟合连接关系，将目标非线性运算函数转换为马赫-曾德尔干涉仪待拟合的拟合非线性运算函数。
实际应用中，因为目标非线性运算函数与MZI的结构并不相适应，所以在基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系之后，还需基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数。
步骤S104：基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数。
实际应用中，因为MZI只能实现线性乘加运算，所以在基于目标光学芯片实现目标非线性运算函数的过程中，还需基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数，此外，因为拟合非线性运算函数可以使得目标非线性运算函数与MZI的结构相适应，所以基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数后，可以保证MZI实现目标非线性运算函数。
步骤S105：基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括马赫-曾德尔干涉仪的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片。
实际应用中，在基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数之后，虽然可以获知拟合线性运算函数及拟合连接关系，但拟合连接关系只是表明了MZI实现目标非线性运算函数所用的结构，该MZI中的配置参数并不确定，所以还需基于拟合线性运算函数及拟合连接关系便可以基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构可以包括马赫-曾德尔干涉仪的数量、连接关系及配置参数，以便基于目标光学芯片的结构设计能够实现目标非线性运算函数的目标光学芯片。
本申请提供的一种光学芯片设计方法，获取目标光学芯片待实现的目标非线性运算函数，目标光学芯片基于MZI设计；基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系；基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数；基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数；基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括MZI的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片。本申请中，可以基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系，使得MZI的拟合连接关系与目标非线性运算函数相适应；可以基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数，使得MZI待实现的拟合非线性运算函数与自身结构相适应；并且可以基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数，基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，使得目标光学芯片的运算特性与MZI的连接架构相适应，保证了目标光学芯片基于MZI实现目标非线性运算函数的准确性，且整个光学芯片设计过程中无需将光信号转换为电信号，结构简便，运算效率高。
本申请实施例提供的一种光学芯片设计方法中，在基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系的过程中，因为目标非线性运算函数的自变量需作为MZI的输入，而 该自变量输入MZI后会按照MZI的结构进行相应运算，且只有运算特征与目标非线性运算函数的特征一致的输出才能作为MZI实现目标非线性运算函数的模拟输出，所以此过程可以包括以下步骤：将目标非线性运算函数中的自变量作为MZI的输入，基于MZI的结构确定MZI两个输出的连接关系；将特征与目标非线性运算函数一致的输出的连接关系，确定为拟合连接关系。
为了便于理解，仍以目标非线性运算函数为sigmoid函数为例，其表达式为：
Figure PCTCN2022074612-appb-000012
因为此时只有一个自变量，所以MZI的一个输入端口需封住，此时MZI的结构如图3所示，且sigmoid函数的自变量输入MZI后的相应输出为：
Figure PCTCN2022074612-appb-000013
因为通过MZI能够观测到的只有该输出中的是不，所以该自变量输入MZI后的相应输出可以等效为
Figure PCTCN2022074612-appb-000014
最后，由sigmoid函数的曲线特征可知，函数值不能小于0，只有MZI中的-xsinθcosθ这种连接关系可以实现拟合，所以sigmoid函数在MZI中的拟合连接关系为：
Figure PCTCN2022074612-appb-000015
此时，在基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数的过程中，因为-xsinθcosθ等效于xsin2θ，该函数会使得自变量的系数变化，所以sigmoid函数对应的拟合非线性运算函数可以为
Figure PCTCN2022074612-appb-000016
本申请实施例提供的一种光学芯片设计方法中，在基于最小二乘法确定拟合非线性运算函数的拟合线性运算函数的过程中，存在目标光学芯片无法完美实现目标非线性运算函数的情况，此时便需要对目标光学芯片实现目标非线性运算函数的精度进行控制，相应的，可以根据该精度来确定拟合线性运算函数，也即可以获取目标光学芯片的设计精度；基于设计精度及最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数。
实际应用中，因为目标非线性运算函数是非线性的，而拟合线性运算函数是线性的，所以自变量的取值范围会影响将目标非线性运算函数拟合为拟合线性运算函数的拟合精度，相应的，可以借助自变量的范围来表征设计精度，也即设计精度可以包括自变量的数据集范围；
相应的，在基于设计精度及最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数的过程中，可以基于最小二乘法确定拟合非线性运算函数在各个次方时的初始线性运算函数；在数据集范围内，确定各个初始线性运算函数与拟合非线性运算函数间的相似度；将相似度值最大的初始线性运算函数作为拟合线性运算函数。
为了便于理解，仍以上述sigmoid函数为例，在数据集[-10,10]内对
Figure PCTCN2022074612-appb-000017
进行最小二乘法线性拟合，假设预设次方为1至5次方，则拟合结果可以为：
Figure PCTCN2022074612-appb-000018
将各个拟合结果与sigmoid函数进行相似性比较，比较结果如图4所示，由此可知，最合适拟合的次方为3，所以sigmoid函数的拟合线性运算函数可以为：f(x)≈0.0018x 3-0.2533x-1 f(x)≈0.0018x 3-0.2533x-1。
本申请实施例提供的一种光学芯片设计方法中，在基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构的过程中，因为在MZI的应用过程中，可以在光域使用奇异值分解(Singular Value Decomposition，SVD)的方式实现任意维度的矩阵乘法运算，所以本申请中可以基于拟合线性运算函数、拟合连接关系及奇异值分解法，确定目标光学芯片的结构。
实际应用中，在基于拟合线性运算函数、拟合连接关系及奇异值分解法，确定目标光学芯片的结构的过程中，可以基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的链接算术矩阵；基于奇异值分解法对链接算术矩阵进行分解，得到矩阵分解结果；基于矩阵分解结果确定目标光学芯片的结构。
仍以上述sigmoid函数为例，假设sigmoid函数的拟合线性运算函数为：f(x)≈0.0018x 3-0.2533x-1，0.0018x 3为复合结构，可以通过设计其他器件结构来实现，假设其结构如图5所示，x的输出首先经过光电二级管转化为电能，因为电能是基于光能量转换的，此时参数为|x| 2，对其得到的电压进行k倍的放大，再经过电转化为移相控制器的热，控制调相，从而可以得到一个从|x| 2到相位角θ的调整关系如下：k *|x| 2＝2θ；进一步的，图5中输出的运算关系为x *sin(k|x| 2)，基于三角函数运算关系可以得到x *sin(k|x| 2)≈kx *|x| 2＝0.0018x 3，由此可知第一放大器的放大倍数可以为k＝0.0018，所以目标光学芯片最终的映射关系可以为f(x)≈x sin(0.0018|x| 2-0.2533x-1；因为目标光学芯片最终的映射关系中的-0.2533x可以轻易通过配置θ＝(arcsin(-0.2533))/2在图3的MZI结构中实现，目标光学芯片最终的映射关系中的-1可以通过引入固定光信息通过MZI进行SVD分解后的标准乘加运算实现；而目标光学芯片最终的映射关系中的x sin(0.0018|x| 2)已经通过图5所示结构实现，所以目标光学芯片的链接算术矩阵可以为：
A＝[1 -0.2533 -1]；
对其进行SVD分别后，可得：
Figure PCTCN2022074612-appb-000019
假设
Figure PCTCN2022074612-appb-000020
由现有基于MZI设计的目标拓扑结构实现，则目标光学芯片的结构可以如图6所示，可以包括：第一MZI11、光电二极管12、第一放大器13、 移相控制器14、第二MZI15、目标拓扑结构16；第一MZI的第一光输入端与第二光输入端相连，第一MZI的第二光输出端悬空，第一MZI的第一光输出端与光电二极管的输入端、第二MZI的第一光输入端及目标拓扑结构的第二光输入端分别相连；光电二极管的输出端与第一放大器相连；第一放大器与移相控制器相连；移相控制器与第二MZI的移相器相连；第二MZI的第二光输入端及第二光输出端均悬空；第二MZI的第一光输出端与目标拓扑结构的第一光输入端连接；其中，目标拓扑结构的类型可以包括但不限于：Gridnet拓扑结构、Reck拓扑结构等，比如目标拓扑结构可以包括第三MZI161，第四MZI162，第五MZI163，激光发射器，此时，目标拓扑结构的第一光输入端为第三MZI的第一光输入端，目标拓扑结构的第二光输入端为第三MZI的第二光输入端，目标拓扑结构的光输出端为第五MZI的第一光输出端；第三MZI的第一光输出端与第五MZI的第一光输入端连接，第三MZI的第二光输出端与第四MZI的第一光输入端连接；激光发射器与第四MZI的第二光输入端连接；第四MZI的第一光输出端与第五MZI的第二光输入端连接，第四MZI的第二光输出端悬空；第五MZI的第二光输出端悬空。
请参阅图7，图7为本申请实施例提供的一种光线芯片设计系统的结构示意图。
本申请实施例提供的一种光学芯片设计系统，可以包括：
第一获取模块101，用于获取目标光学芯片待实现的目标非线性运算函数，目标光学芯片基于MZI设计；
第一确定模块102，用于基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系；
第一转换模块103，用于基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数；
第二确定模块104，用于基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数；
第三确定模块105，用于基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括MZI的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片。
本申请实施例提供的一种光学芯片设计系统，第一确定模块可以包括：
第一确定单元，用于将目标非线性运算函数中的自变量作为MZI的输入，基于MZI的结构确定MZI两个输出的连接关系；
第二确定单元，用于将特征与目标非线性运算函数一致的输出的连接关系，确定为拟合连接关系。
本申请实施例提供的一种光学芯片设计系统，第二确定模块可以包括：
第一获取单元，用于获取目标光学芯片的设计精度；
第三确定单元，用于基于设计精度及最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数。
本申请实施例提供的一种光学芯片设计系统，设计精度包括自变量的数据集范围；
第三确定单元可以具体用于：基于最小二乘法确定拟合非线性运算函数在各个次方时的初始线性运算函数；在数据集范围内，确定各个初始线性运算函数与拟合非线性运算函数间的相似度；将相似度值最大的初始线性运算函数作为拟合线性运算函数。
本申请实施例提供的一种光学芯片设计系统，第三确定模块可以包括：
第三确定单元，用于基于拟合线性运算函数、拟合连接关系及奇异值分解法，确定目标光学芯片的结构。
本申请实施例提供的一种光学芯片设计系统，第三确定单元可以具体用于：基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的链接算术矩阵；基于奇异值分解法对链接算术矩阵进行分解，得到矩阵分解结果；基于矩阵分解结果确定目标光学芯片的结构。
本申请实施例提供的一种光学芯片设计系统，目标非线性运算函数包括sigmoid函数；目标光学芯片的结构包括：第一MZI、光电二极管、第一放大器、移相控制器、第二MZI、目标拓扑结构；第一MZI的第一光输入端与第二光输入端相连，第一MZI的第二光输出端悬空，第一MZI的第一光输出端与光电二极管的输入端、第二MZI的第一光输入端及目标拓扑结构的第二光输入端分别相连；光电二极管的输出端与第一放大器相连；第一放大器与移相控制器相连；移相控制器与第二MZI的移相器相连；第二MZI的第二光输入端及第二光输出端均悬空；第二MZI的第一光输出端与目标拓扑结构的第一光输入端连接；
其中，目标拓扑结构的类型包括：Gridnet拓扑结构、Reck拓扑结构。
本申请还提供了一种光学芯片设计设备及计算机可读存储介质，其均具有本申请实施例提供的一种光学芯片设计方法具有的对应效果。请参阅图8，图8为本申请实施例提供的一种光学芯片设计设备的结构示意图。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：
获取目标光学芯片待实现的目标非线性运算函数，目标光学芯片基于MZI设计；
基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系；
基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数；
基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数；
基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括MZI的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：将目标非线性运算函数中的自变量作为MZI的输入，基于MZI的结构确定MZI两个输出的连接关系；将特征与目标非线性运算函数一致的输出的连接关系，确定为拟合连接关系。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：获取目标光学芯片的设计精度；基于设计精度及最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：设计精度包括自变量的数据集范围；基于最小二乘法确定拟合非线性运算函数在各个次方时的初始线性运算函数；在数据集范围内，确定各个初始 线性运算函数与拟合非线性运算函数间的相似度；将相似度值最大的初始线性运算函数作为拟合线性运算函数。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：基于拟合线性运算函数、拟合连接关系及奇异值分解法，确定目标光学芯片的结构。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的链接算术矩阵；基于奇异值分解法对链接算术矩阵进行分解，得到矩阵分解结果；基于矩阵分解结果确定目标光学芯片的结构。
本申请实施例提供的一种光学芯片设计设备，包括存储器201和处理器202，存储器201中存储有计算机程序，处理器202执行计算机程序时实现如下步骤：目标非线性运算函数包括sigmoid函数；目标光学芯片的结构包括：第一MZI、光电二极管、第一放大器、移相控制器、第二MZI、目标拓扑结构；第一MZI的第一光输入端与第二光输入端相连，第一MZI的第二光输出端悬空，第一MZI的第一光输出端与光电二极管的输入端、第二MZI的第一光输入端及目标拓扑结构的第二光输入端分别相连；光电二极管的输出端与第一放大器相连；第一放大器与移相控制器相连；移相控制器与第二MZI的移相器相连；第二MZI的第二光输入端及第二光输出端均悬空；第二MZI的第一光输出端与目标拓扑结构的第一光输入端连接；其中，目标拓扑结构的类型包括：Gridnet拓扑结构、Reck拓扑结构。
请参阅图9，本申请实施例提供的另一种光学芯片设计设备中还可以包括：与处理器202连接的输入端口203，用于传输外界输入的命令至处理器202；与处理器202连接的显示单元204，用于显示处理器202的处理结果至外界；与处理器202连接的通信模块205，用于实现光学芯片设计设备与外界的通信。显示单元204可以为显示面板、激光扫描使显示器等；通信模块205所采用的通信方式包括但不局限于移动高清链接技术(HML)、通用串行总线(USB)、高清多媒体接口(HDMI)、无线连接：无线保真技术(WiFi)、蓝牙通信技术、低功耗蓝牙通信技术、基于IEEE802.11s的通信技术。
本申请实施例提供的一种计算机可读存储介质601，请参阅图10，图10为本申请实施例提供的一种计算机可读存储介质的结构示意图，计算机可读存储介质601中存储有计算机程序610，计算机程序610被处理器执行时实现如下步骤：
获取目标光学芯片待实现的目标非线性运算函数，目标光学芯片基于MZI设计；
基于目标非线性运算函数中的自变量及MZI的结构，确定MZI的拟合连接关系；
基于拟合连接关系，将目标非线性运算函数转换为MZI待拟合的拟合非线性运算函数；
基于最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数；
基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的结构，目标光学芯片的结构包括MZI的数量、连接关系及配置参数，以基于目标光学芯片的结构设计目标光学芯片。
本申请实施例提供的一种计算机可读存储介质，计算机可读存储介质中存储有计算机程序，计算机程序被处理器执行时实现如下步骤：将目标非线性运算函数中的自变量作为MZI的输入，基于MZI的结构确定MZI两个输出的连接关系；将特征与目标非线性运算函数一致的输出的连接关系，确定为拟合连接关系。
本申请实施例提供的一种计算机可读存储介质，计算机可读存储介质中存储有计算机程序，计算机程序被处理器执行时实现如下步骤：获取目标光学芯片的设计精度；基于设计精度及最小二乘法确定拟合非线性运算函数对应的拟合线性运算函数。
本申请实施例提供的一种计算机可读存储介质，计算机可读存储介质中存储有计算机程序，计算机程序被处理器执行时实现如下步骤：设计精度包括自变量的数据集范围；基于最小二乘法确定拟合非线性运算函数在各个次方时的初始线性运算函数；在数据集范围内，确定各个初始线性运算函数与拟合非线性运算函数间的相似度；将相似度值最大的初始线性运算函数作为拟合线性运算函数。
本申请实施例提供的一种计算机可读存储介质，计算机可读存储介质中存储有计算机程序，计算机程序被处理器执行时实现如下步骤：基于拟合线性运算函数、拟合连接关系及奇异值分解法，确定目标光学芯片的结构。
本申请实施例提供的一种计算机可读存储介质，计算机可读存储介质中存储有计算机程序，计算机程序被处理器执行时实现如下步骤：基于拟合线性运算函数及拟合连接关系，确定目标光学芯片的链接算术矩阵；基于奇异值分解法对链接算术矩阵进行分解，得到矩阵分解结果；基于矩阵分解结果确定目标光学芯片的结构。
本申请实施例提供的一种计算机可读存储介质，计算机可读存储介质中存储有计算机程序，计算机程序被处理器执行时实现如下步骤：目标非线性运算函数包括sigmoid函数；目标光学芯片的结构包括：第一MZI、光电二极管、第一放大器、移相控制器、第二MZI、目标拓扑结构；第一MZI的第一光输入端与第二光输入端相连，第一MZI的第二光输出端悬空，第一MZI的第一光输出端与光电二极管的输入端、第二MZI的第一光输入端及目标拓扑结构的第二光输入端分别相连；光电二极管的输出端与第一放大器相连；第一放大器与移相控制器相连；移相控制器与第二MZI的移相器相连；第二MZI的第二光输入端及第二光输出端均悬空；第二MZI的第一光输出端与目标拓扑结构的第一光输入端连接；其中，目标拓扑结构的类型包括：Gridnet拓扑结构、Reck拓扑结构。
本申请所涉及的计算机可读存储介质包括随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器、硬盘、可移动磁盘、CD-ROM、或技术领域内所公知的任意其它形式的存储介质。
本申请实施例提供的一种光学芯片设计系统、设备及计算机可读存储介质中相关部分的说明请参见本申请实施例提供的一种光学芯片设计方法中对应部分的详细说明，在此不再赘述。另外，本申请实施例提供的上述技术方案中与现有技术中对应技术方案实现原理一致的部分并未详细说明，以免过多赘述。
还需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
对所公开的实施例的上述说明，使本领域技术人员能够实现或使用本申请。对这些实施例的多种修改对本领域技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本申请的精神或范围的情况下，在其它实施例中实现。因此，本申请将不会被限制于本文所示的这些实施例，而是要符合与 本文所公开的原理和新颖特点相一致的最宽的范围。

Claims (12)

  1. 一种光学芯片设计方法，其特征在于，包括：
    获取目标光学芯片待实现的目标非线性运算函数，所述目标光学芯片基于MZI设计；
    基于所述目标非线性运算函数中的自变量及所述MZI的结构，确定所述MZI的拟合连接关系；
    基于所述拟合连接关系，将所述目标非线性运算函数转换为所述MZI待拟合的拟合非线性运算函数；
    基于最小二乘法确定所述拟合非线性运算函数对应的拟合线性运算函数；
    基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的结构，所述目标光学芯片的结构包括所述MZI的数量、连接关系及配置参数，以基于所述目标光学芯片的结构设计所述目标光学芯片。
  2. 根据权利要求1所述的方法，其特征在于，所述基于所述目标非线性运算函数中的自变量及所述MZI的结构，确定所述MZI的拟合连接关系，包括：
    将所述目标非线性运算函数中的自变量作为所述MZI的输入，基于所述MZI的结构确定所述MZI两个输出的连接关系；
    将特征与所述目标非线性运算函数一致的所述输出的连接关系，确定为所述拟合连接关系。
  3. 根据权利要求2所述的方法，其特征在于，所述基于最小二乘法确定所述拟合非线性运算函数的拟合线性运算函数，包括：
    获取所述目标光学芯片的设计精度；
    基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数。
  4. 根据权利要求3所述的方法，其特征在于，所述设计精度包括所述自变量的数据集范围；
    所述基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数，包括：
    基于所述最小二乘法确定所述拟合非线性运算函数在各个次方时的初始线性运算函数；
    在所述数据集范围内，确定各个所述初始线性运算函数与所述拟合非线性运算函数间的相似度；
    将相似度值最大的所述初始线性运算函数作为所述拟合线性运算函数。
  5. 根据权利要求4所述的方法，其特征在于，所述基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的结构，包括：
    基于所述拟合线性运算函数、所述拟合连接关系及奇异值分解法，确定所述目标光学芯片的结构。
  6. 根据权利要求5所述的方法，其特征在于，所述基于所述拟合线性运算函数、所述拟合连接关系及奇异值分解法，确定所述目标光学芯片的结构，包括：
    基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的链接算术矩阵；
    基于所述奇异值分解法对所述链接算术矩阵进行分解，得到矩阵分解结果；
    基于所述矩阵分解结果确定所述目标光学芯片的结构。
  7. 根据权利要求6所述的方法，其特征在于，所述目标非线性运算函数包括sigmoid函数；
    所述目标光学芯片的结构包括：第一MZI、光电二极管、第一放大器、移相控制器、第二MZI、目标拓扑结构；所述第一MZI的第一光输入端与第二光输入端相连，所述第一MZI的第二光输出端悬空，所述第一MZI的第一光输出端与所述光电二极管的输入端、所述第二MZI的第一光输入端及所述目标拓扑结构的第二光输入端分别相连；所述光电二极管的输出端与所述第一放大器相连；所述第一放 大器与所述移相控制器相连；所述移相控制器与所述第二MZI的移相器相连；所述第二MZI的第二光输入端及第二光输出端均悬空；所述第二MZI的第一光输出端与所述目标拓扑结构的第一光输入端连接。
  8. 一种光学芯片设计系统，其特征在于，包括：
    第一获取模块，用于获取目标光学芯片待实现的目标非线性运算函数，所述目标光学芯片基于MZI设计；
    第一确定模块，用于基于所述目标非线性运算函数中的自变量及所述MZI的结构，确定所述MZI的拟合连接关系；
    第一转换模块，用于基于所述拟合连接关系，将所述目标非线性运算函数转换为所述MZI待拟合的拟合非线性运算函数；
    第二确定模块，用于基于最小二乘法确定所述拟合非线性运算函数对应的拟合线性运算函数；
    第三确定模块，用于基于所述拟合线性运算函数及所述拟合连接关系，确定所述目标光学芯片的结构，所述目标光学芯片的结构包括所述MZI的数量、连接关系及配置参数，以基于所述目标光学芯片的结构设计所述目标光学芯片。
  9. 根据权利要求8所述的系统，其特征在于，所述基于最小二乘法确定所述拟合非线性运算函数的拟合线性运算函数，包括：
    获取所述目标光学芯片的设计精度；
    基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数。
  10. 根据权利要求9所述的系统，其特征在于，所述设计精度包括所述自变量的数据集范围；
    所述基于所述设计精度及所述最小二乘法确定所述拟合非线性运算函数对应的所述拟合线性运算函数，包括：
    基于所述最小二乘法确定所述拟合非线性运算函数在各个次方时的初始线性运算函数；
    在所述数据集范围内，确定各个所述初始线性运算函数与所述拟合非线性运算函数间的相似度；
    将相似度值最大的所述初始线性运算函数作为所述拟合线性运算函数。
  11. 一种光学芯片设计设备，其特征在于，包括：
    存储器，用于存储计算机程序；
    处理器，用于执行所述计算机程序时实现如权利要求1至7任一项所述光学芯片设计方法的步骤。
  12. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质中存储有计算机程序，所述计算机程序被处理器执行时实现如权利要求1至7任一项所述光学芯片设计方法的步骤。
PCT/CN2022/074612 2021-09-23 2022-01-28 一种光学芯片设计方法、系统、设备及计算机存储介质 Ceased WO2023045202A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111110483.7 2021-09-23
CN202111110483.7A CN113569506B (zh) 2021-09-23 2021-09-23 一种光学芯片设计方法、系统、设备及计算机存储介质

Publications (1)

Publication Number Publication Date
WO2023045202A1 true WO2023045202A1 (zh) 2023-03-30

Family

ID=78173979

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/074612 Ceased WO2023045202A1 (zh) 2021-09-23 2022-01-28 一种光学芯片设计方法、系统、设备及计算机存储介质

Country Status (2)

Country Link
CN (1) CN113569506B (zh)
WO (1) WO2023045202A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113569506B (zh) * 2021-09-23 2021-12-17 苏州浪潮智能科技有限公司 一种光学芯片设计方法、系统、设备及计算机存储介质
CN114567396B (zh) * 2022-02-28 2024-09-06 哲库科技(北京)有限公司 无线通信方法、非线性函数的拟合方法、终端及设备
CN115526291A (zh) * 2022-09-16 2022-12-27 山东云海国创云计算装备产业创新中心有限公司 一种池化结构设计方法、装置、电子设备及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110197277A (zh) * 2019-05-13 2019-09-03 浙江大学 实现数字识别的光学神经网络方法
CN111563583A (zh) * 2020-04-03 2020-08-21 清华大学 基于神经常微分方程的光计算芯片设计方法及光计算芯片
US20200401012A1 (en) * 2019-06-18 2020-12-24 National University Of Defense Technology Reconfigurable nonlinear frequency conversion waveguide chip based on Mach-Zehnder interferometer coupled microring
US20210072784A1 (en) * 2020-11-17 2021-03-11 Intel Corporation Optical nonlinearity and amplification devices for optical neural networks
CN113325650A (zh) * 2021-05-28 2021-08-31 山东云海国创云计算装备产业创新中心有限公司 一种光学电路、光信号处理方法、装置及可读存储介质
CN113343814A (zh) * 2021-05-31 2021-09-03 太原理工大学 一种基于单节点光子储备池计算的手写数字图像识别方法
CN113569506A (zh) * 2021-09-23 2021-10-29 苏州浪潮智能科技有限公司 一种光学芯片设计方法、系统、设备及计算机存储介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103852164B (zh) * 2012-11-30 2016-03-23 中国科学院电子学研究所 一种获取光源光谱的方法
US10133149B2 (en) * 2015-11-25 2018-11-20 Massachusetts Institute Of Technology Apparatus and methods for generating nonlinear effects in centrosymmetric materials
CA3061993C (en) * 2017-05-22 2025-04-01 Imec Vzw III-V/IV LASER SENSOR SYSTEM-CHIP FOR REAL-TIME MONITORING OF BLOOD COMPONENT CONCENTRATION LEVELS

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110197277A (zh) * 2019-05-13 2019-09-03 浙江大学 实现数字识别的光学神经网络方法
US20200401012A1 (en) * 2019-06-18 2020-12-24 National University Of Defense Technology Reconfigurable nonlinear frequency conversion waveguide chip based on Mach-Zehnder interferometer coupled microring
CN111563583A (zh) * 2020-04-03 2020-08-21 清华大学 基于神经常微分方程的光计算芯片设计方法及光计算芯片
US20210072784A1 (en) * 2020-11-17 2021-03-11 Intel Corporation Optical nonlinearity and amplification devices for optical neural networks
CN113325650A (zh) * 2021-05-28 2021-08-31 山东云海国创云计算装备产业创新中心有限公司 一种光学电路、光信号处理方法、装置及可读存储介质
CN113343814A (zh) * 2021-05-31 2021-09-03 太原理工大学 一种基于单节点光子储备池计算的手写数字图像识别方法
CN113569506A (zh) * 2021-09-23 2021-10-29 苏州浪潮智能科技有限公司 一种光学芯片设计方法、系统、设备及计算机存储介质

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"Master's Thesis", 18 May 2018, UNIVERSITY OF ELECTRONIC SCIENCE AND TECHNOLOGY OF CHINA, China, article HOU, JIANHONG: "Design and Packaging on Silicon-based Mach-Zehnder Optical Modulation Chips", pages: 1 - 53, XP009544875 *
SUN, YING: "Research on the Frequncy Modulation Non-linear Correction of Semiconductor Lasers Based on FPGA", CHINESE MASTER’S THESES FULL-TEXT DATABASE, 1 June 2017 (2017-06-01), China, pages 1 - 72, XP009544874 *

Also Published As

Publication number Publication date
CN113569506A (zh) 2021-10-29
CN113569506B (zh) 2021-12-17

Similar Documents

Publication Publication Date Title
CN111753977B (zh) 光学神经网络卷积层芯片、卷积计算方法和电子设备
WO2023045202A1 (zh) 一种光学芯片设计方法、系统、设备及计算机存储介质
CN113298246B (zh) 数据处理方法、装置及计算机可读存储介质
US12001937B2 (en) Optical neural network, data processing method and apparatus based on same, and storage medium
CN113627605B (zh) 光子神经网络的光学干涉单元配置方法、装置及存储介质
CN104898306A (zh) Mz调制器任意点偏置控制装置及方法
CN115905792A (zh) 一种用于光学实数矩阵计算的马赫曾德尔干涉仪网络
CN114399038A (zh) 一种光学神经网络训练方法、装置、设备及介质
WO2023005084A1 (zh) 光学电路搭建方法、光学电路、光信号处理方法及装置
WO2023061067A1 (zh) 偏置点控制装置和方法、光调制系统及控制设备
Yuan et al. Low-phase quantization error Mach–Zehnder interferometers for high-precision optical neural network training
Lu et al. Optimizing optical neural network design for enhanced compatibility with analog computation
Zheng et al. Neuromorphic network with photonic weighting and photoelectronic nonlinear activation based on SOA and APD
CN113657588B (zh) 一种实现光学神经网络的方法及系统
CN120338011A (zh) 光学神经网络及其实现方法、光学神经网络芯片
US20140314365A1 (en) Electro-optic modulator
CN113325650B (zh) 一种光学电路、光信号处理方法、装置及可读存储介质
CN114548381B (zh) 全光神经网络的激活函数构造方法、装置、设备及介质
Yao et al. The solitary wave solutions to the Klein–Gordon–Zakharov equations by extended rational methods
CN117875376A (zh) 光学神经网络的可变增益电光非线性激活函数的实现方法
Qu et al. Numerical realization of automatically reconfigurable optical waveform generator based on photonic reservoir computer
US20240037382A1 (en) Method and device for implementing average pooling of neural network, and storage medium
Bi et al. All-optical nonlinear activation function based on a Mach-Zehnder modulator with magneto-optical material
Yang et al. Traveling wave electrode design for a LiNbO3 integrated optical switch
CN115378510B (zh) 硅基调制器产生高线性度信号的方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22871269

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22871269

Country of ref document: EP

Kind code of ref document: A1