1. A data processing method based on an optical neural network is characterized in that an optical interference unit of the optical neural network comprises a first interference light path structure, a phase shifter and a second interference light path structure, wherein the first interference light path structure and the second interference light path structure both comprise an internal phase shifter and an optical splitter, and the method comprises the following steps:

respectively acquiring the splitting ratio of the optical splitters of the first interference light path structure and the second interference light path structure;

if the splitting ratio of each optical splitter meets the splitting compensation condition, respectively acquiring initial light information of an input light signal, intermediate input light information at an input port of the phase shifter, intermediate output light information at an output port of the phase shifter and final output light information;

and when the initial light information and the intermediate input light information, and the intermediate output light information and the final output light information both satisfy a preset light splitting condition of the optical neural network, calculating parameters of internal phase shifters of the first interference light path structure and the second interference light path structure, so as to perform data processing by using the optical neural network based on the parameters.

2. The optical neural network-based data processing method according to claim 1, wherein the input optical signals include two input optical signals, and the calculating parameters of the internal phase shifters of the first and second interferometric optical path structures when a preset light splitting condition of the optical neural network is satisfied between the initial optical information and the intermediate input optical information, and between the intermediate output optical information and the final output optical information comprises:

the first interference optical path structure comprises a first optical splitter, a first internal phase shifter and a second optical splitter; the first optical splitter inputs an input optical signal of the optical neural network to the first internal phase shifter, and the second optical splitter inputs an optical signal output from the first internal phase shifter to the phase shifter;

calling a first interference light path optical parameter calculation relation to determine the parameter of the first internal phase shifter; the optical parameter calculation relation of the first interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,inputting optical information for a first input optical signal of the optical neural network at an intermediate input port of the phase shifter,inputting optical information for a second input optical signal of the optical neural network at an intermediate input port of the phase shifter,L ₁for the initial optical information of the first input optical signal,L ₂for the initial optical information of the second input optical signal,r ₁is the reflectivity of the first beam splitter,r ₂is the reflectivity of the second beam splitter,t ₁is an intermediate parameter and，t ₂is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the first internal phase shifter.

3. The optical neural network-based data processing method according to claim 2, wherein calculating parameters of the internal phase shifters of the first and second interferometric optical path structures when a preset light splitting condition of the optical neural network is satisfied between the initial light information and the intermediate input light information, and between the intermediate output light information and the final output light information comprises:

the preset light splitting condition is 50: 50, calculating a first parameter calculation relational expression according to the first interference light path optical parameter calculation relational expression;

determining parameters of the first internal phase shifter according to the first parameter calculation relation; the first parameter calculation relation is as follows:

。

4. the optical neural network-based data processing method according to claim 1, wherein the input optical signals include two input optical signals, and the calculating parameters of the internal phase shifters of the first and second interferometric optical path structures when a preset light splitting condition of the optical neural network is satisfied between the initial optical information and the intermediate input optical information, and between the intermediate output optical information and the final output optical information comprises:

the second interference optical path structure comprises a third optical splitter, a second internal phase shifter and a fourth optical splitter; the third optical splitter inputs an input optical signal of the phase shifter to the second internal phase shifter, and the fourth optical splitter performs optical splitting processing on an optical signal output by the second internal phase shifter and outputs the optical signal;

calling a second interference light path optical parameter calculation relation to determine the parameter of the second internal phase shifter; the optical parameter calculation relation of the second interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,for the final output optical information of the first input optical signal,for the final output optical information of the second input optical signal,L ₃outputting optical information for the first input optical signal at an intermediate of the phase shifter output ports,L ₄outputting optical information for the second input optical signal at an intermediate of the phase shifter output ports,r ₃is the reflectivity of the third beam splitter,r ₄is the reflectivity of the fourth beam splitter,t ₃is an intermediate parameter and，t ₄is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the second internal phase shifter.

5. The optical neural network-based data processing method according to claim 4, wherein calculating parameters of the internal phase shifters of the first and second interferometric optical path structures when a preset light splitting condition of the optical neural network is satisfied between the initial light information and the intermediate input light information, and between the intermediate output light information and the final output light information comprises:

the preset light splitting condition is 50: 50, calculating a second parameter calculation relational expression according to the second interference light path optical parameter calculation relational expression;

determining a parameter of the second internal phase shifter according to the second parameter calculation relation; the second parameter calculation relation is as follows:

。

6. the optical neural network-based data processing method of claim 1, wherein the preset splitting condition is 50: 50, the judging whether the splitting ratio of each splitter meets the splitting compensation condition includes:

judging whether the splitting ratio of each splitter is 15: 85 to 85: 15, or more.

7. A data processing device based on an optical neural network, wherein an optical interference unit of the optical neural network comprises a first interference light path structure, a phase shifter and a second interference light path structure, and the first interference light path structure and the second interference light path structure each comprise an internal phase shifter and a light splitter, the device comprising:

the data preprocessing module comprises a parameter information acquisition unit which is used for respectively acquiring the splitting ratio of the optical splitters of the first interference light path structure and the second interference light path structure;

an optical information obtaining module, configured to obtain initial optical information of an input optical signal, intermediate input optical information at an input port of the phase shifter, intermediate output optical information at an output port of the phase shifter, and final output optical information, respectively, if a splitting ratio of each optical splitter satisfies a splitting compensation condition;

and a calculating module, configured to calculate parameters of the internal phase shifters of the first interference optical path structure and the second interference optical path structure when the initial optical information and the intermediate input optical information, and the intermediate output optical information and the final output optical information both satisfy a preset light splitting condition of the optical neural network, so as to perform data processing by using the optical neural network based on the parameters.

8. An optical neural network-based data processing apparatus, comprising a processor configured to implement the steps of the optical neural network-based data processing method according to any one of claims 1 to 6 when executing a computer program stored in a memory.

9. A computer-readable storage medium, on which an optical neural network-based data processing program is stored, which when executed by a processor, implements the steps of the optical neural network-based data processing method according to any one of claims 1 to 6.

10. An optical neural network comprising a silicon photonic optical circuit and the optical neural network-based data processing apparatus of claim 8;

the silicon photon optical path comprises an optical interference unit, and the optical interference unit comprises a first interference optical path structure, a phase shifter and a second interference optical path structure;

the first interference optical path structure comprises a first internal phase shifter, a first optical splitter and a second optical splitter; the second interference light path structure comprises a second internal phase shifter, a third optical splitter and a fourth optical splitter;

a first input optical signal and a second input optical signal of the optical neural network are input to the first internal phase shifter through the first optical splitter, input to the phase shifter through the first internal phase shifter by the second optical splitter, input to the second internal phase shifter through the phase shifter by the third optical splitter, and output through the second optical splitter by the second internal phase shifter.

Background

With the development of Technology, society has now entered the era of cloud + AI (Artificial Intelligence) +5G (5 th Generation Mobile Communication Technology), and a dedicated chip supporting a large amount of computation is required to meet the computation requirement of cloud + AI + 5G. The chip is one of the greatest inventions of human beings, and is also the foundation and the core of the modern electronic information industry. The technology is small enough for mobile phones, computers and digital cameras and large enough for 5G, Internet of things and cloud computing, and is a continuous breakthrough based on chip technology. The development of the semiconductor lithography process level is a fundamental stone of an electronic computer taking a chip as a core, the current semiconductor lithography manufacturing process is almost the physical limit of the moore's law, as the manufacturing process is smaller and smaller, the transistor unit in the chip is close to the molecular scale, and the bottleneck effect of the semiconductor manufacturing process is more and more obvious. With the high-speed development of globalization and science and technology, the amount of data to be processed is increased rapidly, corresponding data processing models and algorithms are also increased continuously, and the requirements on computing power and power consumption are increased continuously. However, the existing electronic computers of von neumann architecture and harvard architecture have the problems of transmission bottleneck, power consumption increase, computing power bottleneck and the like, and it is increasingly difficult to meet the requirements of computing power and power consumption in big data era, for example, the artificial intelligence computing requirement is extremely not matched with the traditional chip computing power growth curve, so that the problem of increasing the computing speed and reducing the computing power consumption is the current critical problem. The photon computing method is one of the potential ways to solve the problems of moore's law predicament and von neumann architecture, i.e. the current computational power and power consumption. Photons have the characteristics of light velocity propagation, electromagnetic interference resistance, arbitrary superposition and the like, and compared with electric calculation, the electric calculation has many advantages, such as: the optical signal is transmitted at the speed of light, so that the speed is greatly improved; the light has natural parallel processing capability and mature wavelength division multiplexing technology, has extremely high operation speed and is very suitable for parallel operation, thereby greatly improving the data processing capability, capacity and bandwidth; the optical computing power consumption is expected to be as low as 10-18J/bit, and under the same power consumption, a photonic device is hundreds of times faster than an electronic device, so that an integrated photonic chip with deep learning capability, high computing power and low power consumption is widely applied, for example, in distance measurement, speed measurement and high-resolution imaging laser radars of long-distance and high-speed moving targets, and in novel computing microscopic associated imaging equipment for realizing high-resolution nondestructive detection of internal structures of biological medicines, nano devices and the like.

In recent years, with the gradual failure of moore's law and the increasing requirements of the big data era on the power consumption and speed of the computing system, the characteristics of high speed and low power consumption of the optical computing technology are more and more emphasized. In addition, due to the parallelism operation characteristic of the optical computing technology and the development of algorithms and hardware architectures such as an optical neural network and the like, the most potential solution is provided for the demands of artificial intelligence technologies such as image recognition, voice recognition, virtual reality and the like on computing power. The light calculation can be divided into an analog light calculation and a digital light calculation. The most typical example of the analog light calculation is fourier operation, and fourier transform related calculation, such as convolution calculation, needs to be applied in the field of image processing and the like. The calculation of the fourier transform with a conventional computer is very computationally expensive, and the passage of light through the lens is itself a fourier transform process, which requires almost no time at all. The digital optical calculation is to form a classic logic gate by combining light and an optical device, construct a calculation system similar to the traditional digital electronic calculation principle, and realize calculation through complex logic gate combination operation.

Photon operation of MZI (Mach-Zehnder interferometer) is the most common industrial solution in the aspect of optical neural networks nowadays, one of the research hotspots of ONN (optical neural networks) is to implement optical linear operation based on MZI, many optical linear modules such as matrix-vector multiplication and convolution can be implemented based on the topological structure of MZI, and the splitting ratio of the optical splitter used in classical topological structures such as GridNet and FFTNet is 50: 50. however, because errors are introduced in the device manufacturing process, the splitting ratio of the actually manufactured MZI is not exactly 50: 50, which are highly likely to be biased, can significantly affect the performance of the MZI-based optical neural network, and thus the data processing accuracy.

In view of this, how to solve the current situation that the performance of the optical neural network is not good due to the low light splitting precision of the light splitter, and improve the data processing precision is a technical problem to be solved by those skilled in the art.

Disclosure of Invention

The application provides a data processing method and device based on an optical neural network, a computer readable storage medium and the optical neural network, and the optical splitter of the optical neural network is compensated, so that the technical problem that the performance of the optical neural network is not good due to low splitting precision of the optical splitter in the related art is solved, the performance of the optical neural network is effectively improved, and the data processing precision can be improved.

In order to solve the above technical problems, embodiments of the present invention provide the following technical solutions:

an embodiment of the present invention provides a data processing method based on an optical neural network, where an optical interference unit of the optical neural network includes a first interference light path structure, a phase shifter, and a second interference light path structure, where the first interference light path structure and the second interference light path structure each include an internal phase shifter and an optical splitter, and the method includes:

respectively acquiring initial light information of an input light signal, intermediate input light information at an input port of the phase shifter, intermediate output light information at an output port of the phase shifter and final output light information;

and when the initial light information and the intermediate input light information, and the intermediate output light information and the final output light information both satisfy a preset light splitting condition of the optical neural network, calculating parameters of internal phase shifters of the first interference light path structure and the second interference light path structure, so as to perform data processing by using the optical neural network based on the parameters.

Optionally, the calculating the parameters of the internal phase shifters of the first interference optical path structure and the second interference optical path structure when the initial optical information and the intermediate input optical information and the intermediate output optical information and the final output optical information both satisfy a preset light splitting condition of the optical neural network includes:

the first interference optical path structure comprises a first optical splitter, a first internal phase shifter and a second optical splitter; the first optical splitter inputs an input optical signal of the optical neural network to the first internal phase shifter, and the second optical splitter inputs an optical signal output from the first internal phase shifter to the phase shifter;

calling a first interference light path optical parameter calculation relation to determine the parameter of the first internal phase shifter; the optical parameter calculation relation of the first interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,inputting optical information for a first input optical signal of the optical neural network at an intermediate input port of the phase shifter,inputting optical information for a second input optical signal of the optical neural network at an intermediate input port of the phase shifter,L ₁for the initial optical information of the first input optical signal,L ₂for the initial optical information of the second input optical signal,r ₁is the reflectivity of the first beam splitter,r ₂is the reflectivity of the second beam splitter,t ₁is an intermediate parameter and，t ₂is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the first internal phase shifter.

Optionally, when the initial light information and the intermediate input light information, and the intermediate output light information and the final output light information both satisfy a preset light splitting condition of the optical neural network, calculating parameters of the internal phase shifters of the first interference light path structure and the second interference light path structure includes:

the preset light splitting condition is 50: 50, calculating a first parameter calculation relational expression according to the first interference light path optical parameter calculation relational expression;

determining parameters of the first internal phase shifter according to the first parameter calculation relation; the first parameter calculation relation is as follows:

。

optionally, the calculating the parameters of the internal phase shifters of the first interference optical path structure and the second interference optical path structure when the initial optical information and the intermediate input optical information and the intermediate output optical information and the final output optical information both satisfy a preset light splitting condition of the optical neural network includes:

the second interference optical path structure comprises a third optical splitter, a second internal phase shifter and a fourth optical splitter; the third optical splitter inputs an input optical signal of the phase shifter to the second internal phase shifter, and the fourth optical splitter performs optical splitting processing on an optical signal output by the second internal phase shifter and outputs the optical signal;

calling a second interference light path optical parameter calculation relation to determine the parameter of the second internal phase shifter; the optical parameter calculation relation of the second interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,for the final output optical information of the first input optical signal,for the final output optical information of the second input optical signal,L ₃outputting optical information for the first input optical signal at an intermediate of the phase shifter output ports,L ₄outputting optical information for the second input optical signal at an intermediate of the phase shifter output ports,r ₃is the reflectivity of the third beam splitter,r ₄is the reflectivity of the fourth beam splitter,t ₃is an intermediate parameter and，t ₄is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the second internal phase shifter.

Optionally, when the initial light information and the intermediate input light information, and the intermediate output light information and the final output light information both satisfy a preset light splitting condition of the optical neural network, calculating parameters of the internal phase shifters of the first interference light path structure and the second interference light path structure includes:

the preset light splitting condition is 50: 50, calculating a second parameter calculation relational expression according to the second interference light path optical parameter calculation relational expression;

determining a parameter of the second internal phase shifter according to the second parameter calculation relation; the second parameter calculation relation is as follows:

。

optionally, before respectively obtaining initial optical information of an input optical signal, intermediate input optical information at the input port of the phase shifter, intermediate output optical information at the output port of the phase shifter, and final output optical information, the method further includes:

respectively acquiring the splitting ratio of the optical splitters of the first interference light path structure and the second interference light path structure;

judging whether the splitting ratio of each splitter meets the splitting compensation condition; if yes, the steps of respectively obtaining initial light information of the input light signal, intermediate input light information at the input port of the phase shifter, intermediate output light information at the output port of the phase shifter and final output light information are executed.

Optionally, the preset spectroscopic condition is 50: 50, the judging whether the splitting ratio of each splitter meets the splitting compensation condition includes:

judging whether the splitting ratio of each splitter is 15: 85 to 85: 15, or more.

Another aspect of the embodiments of the present invention provides a data processing apparatus based on an optical neural network, where an optical interference unit of the optical neural network includes a first interference optical path structure, a phase shifter, and a second interference optical path structure, where the first interference optical path structure and the second interference optical path structure each include an internal phase shifter and a beam splitter, and the apparatus includes:

an optical information obtaining module, configured to obtain initial optical information of an input optical signal, intermediate input optical information at an input port of the phase shifter, intermediate output optical information at an output port of the phase shifter, and final output optical information, respectively;

and a calculating module, configured to calculate parameters of the internal phase shifters of the first interference optical path structure and the second interference optical path structure when the initial optical information and the intermediate input optical information, and the intermediate output optical information and the final output optical information both satisfy a preset light splitting condition of the optical neural network, so as to perform data processing by using the optical neural network based on the parameters.

An embodiment of the present invention further provides an optical neural network-based data processing apparatus, including a processor, where the processor is configured to implement the steps of the optical neural network-based data processing method according to any one of the foregoing embodiments when executing a computer program stored in a memory.

Finally, an embodiment of the present invention provides a computer-readable storage medium, on which a data processing program based on an optical neural network is stored, and when being executed by a processor, the data processing program based on the optical neural network implements the steps of the data processing method based on the optical neural network according to any one of the previous items.

The technical scheme that the application provided's advantage lies in, adopt two to interfere the optical path structure and replace the optical splitter in the silicon optical subset integrated circuit of former optical neural network, and compensate wanting beam splitting ratio with arbitrary optical splitter through the parameter of adjusting the inside looks ware that interferes optical path structure and second and interfere the optical path structure, thereby solve in the correlation technique because the beam splitting precision of optical splitter leads to the not good technical problem of optical neural network performance, effectively promote optical neural network performance, the optical neural network that the utilization was compensated to the optical splitter carries out data processing, realize various complicacies and accurate linear optical module, and then can promote the data processing precision.

In addition, the embodiment of the invention also provides a corresponding implementation device, a computer readable storage medium and an optical neural network aiming at the data processing method based on the optical neural network, so that the method has higher practicability, and the device, the computer readable storage medium and the optical neural network have corresponding advantages.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the related art, the drawings required to be used in the description of the embodiments or the related art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

Fig. 1 is a schematic structural diagram of a conventional optical neural network according to an embodiment of the present invention;

fig. 2 is a schematic flowchart of a data processing method based on an optical neural network according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of an optical path structure of an optical neural network according to an exemplary embodiment of the present invention;

FIG. 4 is a block diagram of an embodiment of a data processing apparatus based on an optical neural network according to the present invention;

fig. 5 is a block diagram of another embodiment of a data processing apparatus based on an optical neural network according to an embodiment of the present invention.

Detailed Description

In order that those skilled in the art will better understand the disclosure, the invention will be described in further detail with reference to the accompanying drawings and specific embodiments. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The terms "first," "second," "third," "fourth," and the like in the description and claims of this application and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements but may include other steps or elements not expressly listed.

Having described the technical solutions of the embodiments of the present invention, various non-limiting embodiments of the present application are described in detail below.

Referring to fig. 2, fig. 2 is a schematic flow chart of a data processing method based on an optical neural network according to an embodiment of the present invention, where a conventional optical interference unit is an optical structure as shown in fig. 1, that is, the conventional optical interference unit includes two optical splitters and a phase shifter, each input optical signal is sequentially processed by a left optical splitter, the phase shifter and a right optical splitter and then output, generally, because a deviation is introduced in a manufacturing process of the optical splitters, a light split of a target chart is, for example, 50: 50 in the actual use process, the input optical signal is not processed strictly according to the splitting ratio parameter, but the parameter is still used in the calculation, which results in that the calculation result is not matched with the actual effect, and especially when the deviation is large, the calculation performance of the optical neural network is not good. In view of this, in order to solve the current situation, the present application may compensate for the self-bias of the optical splitter, thereby improving the network performance. In order to solve the problem, the optical structure of the optical neural network needs to be changed first, and the structure of the optical neural network to which the embodiment of the present invention is applied is: the optical interference unit of the silicon photonic integrated circuit of the optical neural network comprises a first interference optical path structure, a phase shifter and a second interference optical path structure, wherein the first interference optical path structure and the second interference optical path structure both comprise an internal phase shifter and an optical splitter, the optical neural network can comprise a plurality of input optical signals, for example, two input optical signals can be included, for the convenience of distinction, the optical neural network can be called as a first input optical signal and a second input optical signal, and each input optical signal is output after being sequentially processed by the first interference optical path structure, the phase shifter and the second interference optical path structure. For convenience of distinction, the internal phase shifters of the first and second interference optical path structures may be referred to as a first internal phase shifter, a second internal phase shifter, and the optical splitters of the first and second interference optical path structures may be referred to as a first optical splitter, a second optical splitter, a third optical splitter, a fourth optical splitter, and so on. Based on the specific optical structure of the optical neural network, the data processing process based on the optical neural network of the embodiment may include the following steps:

s201: initial light information of an input light signal, intermediate input light information at an input port of the phase shifter, intermediate output light information at an output port of the phase shifter, and final output light information are respectively obtained.

In this step, there may be multiple input optical signals, and those skilled in the art may set up an optical path and determine the number of input optical signals according to the actual application. The initial optical information is the optical information carried by the optical signal of the input layer of the optical neural network, and the optical information may be, for example, optical power, that is, the information carried by the optical signal before being processed by the components of the optical interference unit. The intermediate input optical information is also an optical signal output after the input optical signal is processed by the first interference optical structure, and the intermediate output optical information is also optical information carried by an optical signal output after the optical signal corresponding to the intermediate input optical information is phase-modulated by using the phase shifter. And the final output optical information is the optical information carried by the optical signal output after being processed by the second interference optical path structure.

S202: and when the initial light information and the intermediate input light information, and the intermediate output light information and the final output light information both meet the preset light splitting condition of the optical neural network, calculating parameters of the internal phase shifters of the first interference light path structure and the second interference light path structure so as to perform data processing by using the optical neural network based on the parameters.

In this embodiment, two interference optical path structures are adopted to replace the original optical splitting device, and the phase shifter of the interference optical path structure is adjusted to compensate the deviation of the optical splitting device, so that the parameters of the phase shifter of each interference optical path structure need to be calculated, and when the parameters of the two phase shifters are exactly the required parameters, the parameters can be exactly between the initial optical information and the intermediate input optical information, and the intermediate output optical information and the final output optical information both satisfy the preset light splitting conditions of the optical neural network, that is, the initial optical information and the intermediate input optical information satisfy the preset light splitting conditions, and the intermediate output optical information and the final output optical information also satisfy the preset light splitting conditions. The preset splitting condition, that is, the splitting ratio that the original splitter is intended to achieve, for example, in a typical MZI structure shown in fig. 1, one MZI consists of two splitters and one phase shifter, and ideally, the splitting ratios of the left splitter and the right splitter of the MZI are both 50: 50, the preset spectroscopic condition in this embodiment is also 50: 50. after the parameters of the internal phase shifters are determined, the two internal phase shifters of the optical neural network are adjusted to the parameter values, and then the optical neural network is used for processing data, so that the problem of inaccuracy caused by the deviation of the light splitting device cannot be caused in the data processing result, and the data processing accuracy is improved.

In the technical scheme provided by the embodiment of the invention, two interference light path structures are adopted to replace a silicon light subset of an original optical neural network to form a light splitter in the circuit, and any light splitter is compensated to a desired light splitting ratio by adjusting parameters of internal phase shifters of a first interference light path structure and a second interference light path structure, so that the technical problem of poor optical neural network performance caused by low light splitting precision of the light splitter in the related technology is solved, the optical neural network performance is effectively improved, various complex and accurate linear optical modules are realized by utilizing the optical neural network for compensating the light splitter to perform data processing, and further the data processing precision can be improved.

It should be noted that, in the present application, there is no strict sequential execution order among the steps, and as long as the logical order is met, the steps may be executed simultaneously or according to a certain preset order, and fig. 2 is only an exemplary manner, and does not represent that only the execution order is the order.

In the foregoing embodiment, the execution steps of how to calculate the parameter of the internal phase shifter of the optical neural network are not limited, and a method for calculating the parameter of the internal phase shifter is provided in this embodiment, taking two input optical output signals as an example, the method may include the following steps:

the first interference optical path structure comprises a first optical splitter, a first internal phase shifter and a second optical splitter; the first optical splitter inputs an input optical signal of the optical neural network to the first internal phase shifter, and the second optical splitter inputs an optical signal output from the first internal phase shifter to the phase shifter; the second interference light path structure comprises a third optical splitter, a second internal phase shifter and a fourth optical splitter; the third optical splitter inputs the input optical signal of the phase shifter to the second internal phase shifter, and the fourth optical splitter performs optical splitting processing on the optical signal output from the second internal phase shifter and outputs the optical signal.

Calling an optical parameter calculation relation of a first interference light path to determine parameters of a first internal phase shifter; the optical parameter calculation relation of the first interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,a first input optical signal of the optical neural network inputs optical information at an intermediate input port of the phase shifter,the second input optical signal of the optical neural network inputs optical information at the middle of the phase shifter input port,L ₁for the initial optical information of the first input optical signal,L ₂for the initial optical information of the second input optical signal,r ₁is the reflectivity of the first beam splitter,r ₂is the reflectivity of the second beam splitter,t ₁is an intermediate parameter and，t ₂is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the first internal phase shifter.

Calling a second interference light path optical parameter calculation relation to determine the parameter of a second internal phase shifter; the optical parameter calculation relation of the second interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,for the final output optical information of the first input optical signal,for the final output optical information of the second input optical signal,L ₃the optical information is output for the first input optical signal in the middle of the phase shifter output port,L ₄the optical information is output for the second input optical signal in the middle of the phase shifter output port,r ₃is the reflectivity of the third beam splitter,r ₄is the reflectivity of the fourth light splitter,t ₃is an intermediate parameter and，t ₄is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the second internal phase shifter.

In order to make the implementation of the technical solution of the present application clearly apparent to those skilled in the art, in the present application, with reference to fig. 3, the preset spectroscopic condition is 50: 50, for example, to illustrate the calculation process of the parameters of the internal phase shifters of the first and second interference optical path structures:

if the preset light splitting condition is 50: 50, for the optical path structure shown in fig. 1, ideally, the splitting ratios of the left splitter and the right splitter of the MZI are both 50: 50, then the following conditions are satisfied:

；（1）

；（2）

in the formula (I), the compound is shown in the specification,P _A is the power of the point a and,P _B is composed of_BThe power of the point or points of the power,P _E is the power of the point E, and,P _F is the power of the point F and,representing the power from point a in point C,for the power from point a in point D,representing the power from point B in point C,for the power from point B in point D,representing the power from point E in point G,for the power from point F in point G,representing the power from point E in point H,is the power from point F in point H,

the present embodiment respectively implements a left optical splitter and a right optical splitter based on the interference optical path structure, as shown in fig. 3There are four beam splitters BS1, BS2, BS3 and BS4, corresponding to a reflectivity ofr ₁、r ₂、r ₃、r ₄The parameters of the three phase shifters areAndby adjustingAndcompensation is performed so that equations (1) and (2) hold.

From the MZI transmission matrix, the transmission matrix in the left dashed box in fig. 3 can be obtained as:

；（3）

wherein the content of the first and second substances,，when the manufacture of the device is completed,r ₁、r ₂、t ₁、t ₂are all known. The relationship of the input port A, B and the output port C, D can be expressed as:

。（4）

whereinL ₁AndL ₂the signals corresponding to the a port and the B port,andsignals corresponding to the C port and the D port are substitutedObtaining:

；（5）

1) to make it possible toSuppose that，L ₂=0, which can be obtained from the euler equation and the trigonometric equation:

；（6）

is composed ofThe sum of the square of the real part and the square of the imaginary part, i.e.:

；（7）

order toCan obtain。（8）

2) To make it possible toSuppose that，L ₂=0, one can obtain:

；（9）

；（10）

order toThe following can be obtained:。（11）

3) to make it possible toSuppose thatL ₁=0，It is possible to obtain:

；（12）

；（13）

order toThe following can be obtained:。（14）

4) to is coming toMake itSuppose thatL ₁=0，It is possible to obtain:

；（15）

；（16）

order toThe following can be obtained:。（17）

based onAndit can be seen that equations 8, 11, 14, and 17 are all equal, which satisfies the energy conservation and symmetry of the MZI.

Similarly, based on the derivation process, the internal phase shifter parameters of the right optical splitter, that is, the second interference optical path structure, can be calculated, since the two are the same except for different parameter subscripts, for the sake of brevity of the description, the derivation process is not repeated here, and the calculation relationship of the internal phase shifter parameters of the second interference optical path structure is as follows:

。（18）

based on the above embodiment, in the practical application process, if the preset splitting ratio is 50: 50, calculating according to the optical parameter calculation relation of the first interference light path to obtain a first parameter calculation relation; determining parameters of the first internal phase shifter according to the first parameter calculation relation; the first parameter calculation relationship is:

。

calculating according to the optical parameter calculation relation of the second interference light path to obtain a second parameter calculation relation; determining parameters of a second internal phase shifter according to the second parameter calculation relation; the second parameter calculation relationship is:

。

it can be understood that the compensation of the optical splitter requires the optical parameters of the original optical splitter, and the more the compensation is satisfactory, the higher the compensation accuracy, and accordingly, the better the performance of the optical neural network. Based on this, before obtaining the initial optical information of the two input optical signals, the intermediate input optical information at the input port of the phase shifter, the intermediate output optical information at the output port of the phase shifter, and the final output optical information, the method may further include:

respectively acquiring the splitting ratio of the optical splitters of the first interference light path structure and the second interference light path structure;

judging whether the splitting ratio of each splitter meets the splitting compensation condition; if yes, the steps of respectively obtaining initial light information of the two input light signals, intermediate input light information at an input port of the phase shifter, intermediate output light information at an output port of the phase shifter and final output light information are executed. And if one of the splitting ratios of the optical splitters does not meet the splitting compensation condition, the optical neural network is not subjected to the compensation processing of the optical splitters.

Wherein, the spectral compensation condition can be calculated based on a preset spectral condition, and the preset spectral condition is 50: and 50, under the condition of spectral compensation, the splitting ratio of the optical splitter is 15: 85 to 85: 15, or more. The specific calculation process may include:

because of the fact thatTherefore, it isThe following are satisfied:

；（19）

based onAndtherefore, the following steps are carried out:

；（20）

by simplifying the above equation 21, it is possible to obtain:

；（21）

when in user _i ²When the value is not less than 0.15, the above formula (21) is always satisfied. Because of the fact thatrAndthas the symmetry oft _i ²Not less than 0.15, therefore the method can compensate 15: 85 to 85: 15, a beam splitter.

From the above, the present embodiment can use a splitting ratio of 15: 85 to 85: MZI in the range of 15 instead of 50: 50 splitter, by adjusting the value of the phase shifter to simulate 50: 50, thereby realizing accurate 50: and the MZI of the 50 optical splitters improves the precision and stability of the optical neural network based on the MZI.

The embodiment of the invention also provides a corresponding device for the data processing method based on the optical neural network, so that the method has higher practicability. Wherein the means can be described separately from the functional module point of view and the hardware point of view. The following describes an optical neural network-based data processing apparatus according to an embodiment of the present invention, and the optical neural network-based data processing apparatus described below and the optical neural network-based data processing method described above may be referred to in correspondence.

Based on the angle of the functional module, referring to fig. 4, fig. 4 is a structural diagram of a data processing apparatus based on an optical neural network according to an embodiment of the present invention in a specific implementation, where the structure of the optical neural network applicable to this embodiment is: its optical interference unit includes first interference light path structure, looks ware and second interference light path structure, and first interference light path structure and second interference light path structure all include inside looks ware and beam splitter, and the device includes:

the optical information obtaining module 401 is configured to obtain initial optical information of an input optical signal, intermediate input optical information at an input port of the phase shifter, intermediate output optical information at an output port of the phase shifter, and final output optical information, respectively.

A calculating module 402, configured to calculate parameters of the internal phase shifters of the first interference optical path structure and the second interference optical path structure when the initial optical information and the intermediate input optical information, and the intermediate output optical information and the final output optical information both satisfy a preset light splitting condition of the optical neural network, so as to perform data processing by using the optical neural network based on the parameters.

Optionally, in some embodiments of this embodiment, the calculating module 402 may include a first optical parameter calculating unit, where the first optical parameter calculating unit is configured to: the input optical signal comprises two input optical signals, and the first interference optical path structure comprises a first optical splitter, a first internal phase shifter and a second optical splitter; the first optical splitter inputs an input optical signal of the optical neural network to the first internal phase shifter, and the second optical splitter inputs an optical signal output from the first internal phase shifter to the phase shifter; calling an optical parameter calculation relation of a first interference light path to determine parameters of a first internal phase shifter; the optical parameter calculation relation of the first interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,a first input optical signal of the optical neural network inputs optical information at an intermediate input port of the phase shifter,the second input optical signal of the optical neural network inputs optical information at the middle of the phase shifter input port,L ₁for the initial optical information of the first input optical signal,L ₂for the initial optical information of the second input optical signal,r ₁is the reflectivity of the first beam splitter,r ₂is the reflectivity of the second beam splitter,t ₁is an intermediate parameter and，t ₂is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the first internal phase shifter.

As an optional implementation manner of this embodiment, the first optical parameter calculating unit may be further configured to: the preset light splitting condition is 50: 50, calculating according to the optical parameter calculation relational expression of the first interference light path to obtain a first parameter calculation relational expression; determining parameters of the first internal phase shifter according to the first parameter calculation relation; the first parameter calculation relationship is:

。

optionally, in another embodiment of this embodiment, the calculating module 402 may include a second optical parameter calculating unit, where the second optical parameter calculating unit is configured to: the input optical signals comprise two paths of input optical signals, and the second interference optical path structure comprises a third optical splitter, a second internal phase shifter and a fourth optical splitter; the third optical splitter inputs the input optical signal of the phase shifter to the second internal phase shifter, and the fourth optical splitter performs optical splitting processing on the optical signal output from the second internal phase shifter and outputs the optical signal; calling a second interference light path optical parameter calculation relation to determine the parameter of a second internal phase shifter; the optical parameter calculation relation of the second interference light path is as follows:

；

in the formula (I), the compound is shown in the specification,for the final output optical information of the first input optical signal,for the final output optical information of the second input optical signal,L ₃the optical information is output for the first input optical signal in the middle of the phase shifter output port,L ₄the optical information is output for the second input optical signal in the middle of the phase shifter output port,r ₃is the reflectivity of the third beam splitter,r ₄is the reflectivity of the fourth light splitter,t ₃is an intermediate parameter and，t ₄is an intermediate parameter and，ithe number of the imaginary numbers is represented,ethe index is expressed as a function of time,is a parameter of the second internal phase shifter.

As an optional implementation manner of this embodiment, the second optical parameter calculating unit is configured to: the preset light splitting condition is 50: 50, calculating according to the optical parameter calculation relational expression of the second interference light path to obtain a second parameter calculation relational expression; determining parameters of a second internal phase shifter according to the second parameter calculation relation; the second parameter calculation relationship is:

。

optionally, in some other embodiments of this embodiment, the apparatus may further include a data preprocessing module, for example, where the data preprocessing module includes:

the condition presetting unit is used for presetting light splitting compensation conditions;

the parameter information acquisition unit is used for respectively acquiring the splitting ratio of the optical splitters of the first interference light path structure and the second interference light path structure;

the judgment execution module is used for judging whether the splitting ratio of each optical splitter meets the splitting compensation condition; if yes, the optical information obtaining module 401 continues to be executed.

As an optional implementation manner of this embodiment, the condition presetting unit may be configured to, if the preset splitting condition is 50: 50, setting the splitting compensation condition in advance that the splitting ratio of the splitter is 15: 85 to 85: 15, in the range of the first and second cells.

In the embodiment of the present invention, the functions of each functional module of the data processing apparatus based on the optical neural network may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the related description of the foregoing method embodiment, which is not described herein again.

Therefore, the embodiment of the invention can solve the technical problem of poor performance of the optical neural network due to low light splitting precision of the light splitter in the related art by compensating the light splitter of the optical neural network, effectively improve the performance of the optical neural network and further improve the data processing precision.

The above mentioned optical neural network based data processing device is described from the perspective of functional modules, and further, the present application also provides an optical neural network based data processing device, which is described from the perspective of hardware. Fig. 5 is a block diagram of another optical neural network-based data processing apparatus according to an embodiment of the present disclosure. As shown in fig. 5, the apparatus comprises a memory 50 for storing a computer program; a processor 51 for implementing the steps of the optical neural network-based data processing method as mentioned in any of the above embodiments when executing the computer program.

The processor 51 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 51 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 51 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 51 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, the processor 51 may further include an AI (Artificial Intelligence) processor for processing computing operations related to machine learning.

Memory 50 may include one or more computer-readable storage media, which may be non-transitory. Memory 50 may also include high speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In this embodiment, the memory 50 is at least used for storing a computer program 501, wherein the computer program can realize the relevant steps of the optical neural network-based data processing method disclosed in any one of the foregoing embodiments after being loaded and executed by the processor 51. In addition, the resources stored in the memory 50 may also include an operating system 502, data 503, and the like, and the storage manner may be a transient storage manner or a permanent storage manner. Operating system 502 may include Windows, Unix, Linux, etc. The data 503 may include, but is not limited to, data corresponding to a data processing result based on an optical neural network, and the like.

In some embodiments, the optical neural network-based data processing apparatus may further include a display screen 52, an input/output interface 53, a communication interface 54 or network interface, a power supply 55, and a communication bus 56. The display 52 and the input/output interface 53, such as a Keyboard (Keyboard), belong to a user interface, and the optional user interface may also include a standard wired interface, a wireless interface, and the like. Alternatively, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, or the like. The display, which may also be referred to as a display screen or display unit, is suitable for displaying information processed in the optical neural network-based data processing apparatus and for displaying a visualized user interface. The communication interface 54 may optionally include a wired interface and/or a wireless interface, such as a WI-FI interface, a bluetooth interface, etc., which are typically used to establish a communication link between the optical neural network-based data processing apparatus and other electronic devices. The communication bus 56 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown in FIG. 5, but this is not intended to represent only one bus or type of bus.

Those skilled in the art will appreciate that the architecture shown in fig. 5 does not constitute a limitation of an optical neural network-based data processing apparatus, and may include more or fewer components than those shown, for example, and may also include sensors 57 that perform various functions.

The functions of the functional modules of the data processing apparatus based on the optical neural network according to the embodiment of the present invention may be specifically implemented according to the method in the above method embodiment, and the specific implementation process may refer to the related description of the above method embodiment, which is not described herein again.

Therefore, the embodiment of the invention can solve the technical problem of poor performance of the optical neural network due to low light splitting precision of the light splitter in the related art by compensating the light splitter of the optical neural network, effectively improve the performance of the optical neural network and further improve the data processing precision.

It is to be understood that, if the data processing method based on the optical neural network in the above embodiments is implemented in the form of software functional units and sold or used as a stand-alone product, it may be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application may be substantially or partially implemented in the form of a software product, which is stored in a storage medium and executes all or part of the steps of the methods of the embodiments of the present application, or all or part of the technical solutions. And the aforementioned storage medium includes: a U disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), an electrically erasable programmable ROM, a register, a hard disk, a removable magnetic disk, a CD-ROM, a magnetic or optical disk, and other various media capable of storing program codes.

Based on this, the embodiment of the present invention further provides a computer readable storage medium, which stores a data processing program based on an optical neural network, and the data processing program based on the optical neural network is executed by a processor, and the steps of the data processing method based on the optical neural network are as described in any one of the above embodiments.

Finally, an embodiment of the present invention further provides an optical neural network, which may include the following contents:

it is understood that the optical neural network may comprise an input layer, an output layer and intermediate hidden layers, each intermediate hidden layer comprising an Optical Interference Unit (OIU) and an optical non-linear unit (ONU) acting as matrix multiplication and activation functions, respectively. The optical interference unit may be implemented, for example, by a programmable nanophotonic device based on a Mach-Zehnder interferometer (MZI), which may be formed, for example, by connecting an upper and a lower silicon waveguide branches by a coupler. The inner phase shifter controls the output splitting ratio by changing the refractive index of the waveguide, and the outer phase shifter controls differential output and phase delay. The ONU may be implemented by optical hardware with nonlinear characteristics such as saturable absorbers, optical bistability, etc. Optical Interference Units (OIUs) are integrated in silicon photonics circuits, also known as silicon photonics integrated circuits. The optical neural network of the present embodiment comprises a silicon photonic circuit and the data processing apparatus based on the optical neural network as described in any of the previous embodiments.

The optical interference unit of the silicon photon optical path 61 includes a first interference optical path structure, a phase shifter and a second interference optical path structure. The first interference optical path structure comprises a first internal phase shifter, a first optical splitter and a second optical splitter; the second interference light path structure comprises a second internal phase shifter, a third optical splitter and a fourth optical splitter; a first input optical signal and a second input optical signal of the optical neural network are respectively input into the first internal phase shifter through the first optical splitter, input into the phase shifter through the first internal phase shifter through the second optical splitter, input into the second internal phase shifter through the third optical splitter, and output through the fourth optical splitter through the second internal phase shifter.

Both the first interference optical path and the second interference optical path may be MZI structures, as shown in fig. 3, and of course, other interference optical paths may also be adopted, which do not affect the implementation of the present application.

The functions of each functional module or each optical device and electrical device of the optical neural network according to the embodiments of the present invention may be specifically implemented according to the method in the above method embodiments, and the specific implementation process may refer to the description related to the above method embodiments, and will not be described herein again.

Therefore, the embodiment of the invention can solve the technical problem of poor performance of the optical neural network due to low light splitting precision of the light splitter in the related art by compensating the light splitter of the optical neural network, effectively improve the performance of the optical neural network and further improve the data processing precision.

The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.

Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The data processing method and apparatus based on the optical neural network, the computer readable storage medium and the optical neural network provided by the present application are described in detail above. The principles and embodiments of the present invention are explained herein using specific examples, which are presented only to assist in understanding the method and its core concepts. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present application.