CN113722667A - Data processing method and device based on Italian machine and Italian machine - Google Patents
Data processing method and device based on Italian machine and Italian machine Download PDFInfo
- Publication number
- CN113722667A CN113722667A CN202110797381.0A CN202110797381A CN113722667A CN 113722667 A CN113722667 A CN 113722667A CN 202110797381 A CN202110797381 A CN 202110797381A CN 113722667 A CN113722667 A CN 113722667A
- Authority
- CN
- China
- Prior art keywords
- matrix
- sampling
- spin
- output
- round
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Optical Modulation, Optical Deflection, Nonlinear Optics, Optical Demodulation, Optical Logic Elements (AREA)
Abstract
本发明提供了一种基于伊辛机的数据处理方法、装置及伊辛机，该方法包括：对自旋组态进行循环更新，将更新后的自旋组态调制到高斯光束的相位上，获得输入矩阵；根据输入矩阵与变换矩阵获得输出矩阵；根据输出矩阵确定当前采样轮次的输出光强，根据输出光强和模型特征值矩阵确定哈密顿量，根据哈密顿量确定采样结果，并在确定末轮采样时，使末轮采样对应的自旋组态作为数据处理结果。本发明提供的一种基于伊辛机的数据处理方法、装置及伊辛机，能够将伊辛模型的数据处理过程在光束上完成，并能够实现由光信号到电信号的转换，具有以光速进行信息的并行处理的能力，可以极大提高求解伊辛问题的速率。
The invention provides a data processing method and device based on the Ising machine, and the Ising machine. The method includes: cyclically updating the spin configuration, modulating the updated spin configuration to the phase of the Gaussian beam, Obtain the input matrix; obtain the output matrix according to the input matrix and the transformation matrix; determine the output light intensity of the current sampling round according to the output matrix, determine the Hamiltonian according to the output light intensity and the model eigenvalue matrix, and determine the sampling result according to the Hamiltonian, and When determining the last round of sampling, the spin configuration corresponding to the last round of sampling is used as the data processing result. The data processing method and device based on the Ising machine and the Ising machine provided by the present invention can complete the data processing process of the Ising model on the light beam, and can realize the conversion from optical signal to electrical signal, and has the advantages of speed of light. The ability to perform parallel processing of information can greatly increase the rate at which the Ising problem can be solved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a data processing method and device based on an Italian machine and the Italian machine.
Background
The Esinon model may be used to solve combinatorial optimization problems. The combinatorial optimization problem refers to a class of problems that cannot find a globally optimal solution within polynomial time at present, such as a traveler problem and a maximum segmentation problem.
The IshIn model (Ising model) can be described by a lattice, each point on the lattice has a spin σ, the value of which can only be +1 or-1, and the set of spins of all points is called the spin configuration σ. Introducing an interaction coefficient J for two adjacent points i and Jij. If the interaction of the applied magnetic field h with each spin is reconsidered, the Hamiltonian of the system is defined as:
the solution to the Exin model may be equivalent to solving a corresponding Exin model at JijAnd when the system is unchanged, taking a global energy minimum value, namely the system spin configuration sigma of the energy ground state.
However, the current algorithm needs to perform a large amount of calculation solution, so that the solution process is physically accelerated by light, and such a scheme for accelerating the solution of the izod model in a physical manner is called an izod machine (Ising machine).
There are three main schemes for the present yixinji. The first is an architecture based on optical parametric oscillation and FPGA feedback, in which the ising spin is modulated on the phase of each pulse in the resonator and the FPGA computes the feedback signal to determine the next sampling round. The second is an architecture based on a wavefront modulator where the ising spin is modulated on the phase modulation of each pixel on the wavefront modulator. The third is an on-chip optical matrix transformation architecture based on the Reck structure, in which the ising spin is modulated on the amplitude of the electric field in the waveguide and the next round of feedback calculations is performed in the electrical domain.
For the architecture based on optical parametric oscillation, firstly, the FPGA can only modulate a single pulse in a resonant cavity at the same time, and the time consumption is linearly increased along with the increase of the problem scale; the other is that the FPGA performs operations in the electrical domain, which also weakens the advantages of optical computation to some extent.
For the existing architecture based on the wavefront modulator, the limitation is that any Esin model cannot be mapped onto the parameters of the architecture.
For the Itanium machine based on the on-chip Reck structure, the problem is that the complexity of the system is O (N) along with the increase of the number of spins2) This, in turn, limits the ability of the system to solve the problem with more spin numbers. Meanwhile, the scheme needs to measure the amplitude and the phase of the electric field in the waveguide at the same time, and further increases the complexity of the scheme.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a data processing method and device based on an Itanium machine and the Itanium machine.
The invention provides a data processing method based on an Itanium machine, which comprises the following steps:
circularly updating the spin configuration, and modulating the updated spin configuration to the phase of the Gaussian beam to obtain an input matrix; the spin configuration is generated randomly based on the constructed Esin model;
obtaining an output matrix according to the input matrix and the transformation matrix; wherein the transformation matrix is initially determined based on the constructed Exin model;
and determining the output light intensity of the current sampling round according to the output matrix, determining a Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining a sampling result according to the Hamilton quantity, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
According to the data processing method based on the Itanium machine, the updated spin configuration is modulated on the phase of the Gaussian beam to obtain an input matrix, and the method comprises the following steps:
modulating the updated spin configuration and the electrical feedback signal to the phase of the Gaussian beam, wherein the phase difference pi between the beam representing the +1 spin and the beam representing the-1 spin is the same, and the electric field amplitude of each beam is the same;
the complex amplitudes of each beam are arranged to obtain an input matrix.
The invention provides a data processing method based on an Itanium machine, wherein the obtaining of an output matrix according to an input matrix and a transformation matrix comprises the following steps:
and multiplying the input matrix and the transformation matrix to obtain an output matrix.
The invention provides a data processing method based on an Isci machine, wherein the method comprises the following steps of determining the output light intensity of the current sampling round according to the output matrix and determining the Hamiltonian according to the output light intensity and a model eigenvalue matrix, and comprises the following steps:
determining the output light intensity of the current sampling round by adopting a first calculation formula according to the output matrix;
determining a Hamiltonian quantity by adopting a second calculation formula according to the output light intensity and the model eigenvalue matrix;
wherein the first calculation formula includes:
I＝|AEin|2
where I is the output intensity, A is the transformation matrix, EinTo input a matrix, AEinIs an output matrix; l. capillary2Representing the square of the absolute value of each element in the matrix;
the second calculation formula includes:
H(σ)＝-∑j(Iλ)jjIj
wherein σ represents the spin vector of the Esin model, H (σ) is the Hamiltonian, and IλIs a model eigenvalue matrix, (I)λ)jjIs IλJ-th diagonal element of (1)jIs the jth element of I.
The invention provides a data processing method based on an Italian machine, which further comprises the following steps:
the determining a sampling result according to the Hamilton quantity, and when determining the last round of sampling, using the spin configuration corresponding to the last round of sampling as a data processing result, including:
determining the variation between the Hamiltonian of the current sampling round and the Hamiltonian of the previous sampling round;
if the variable quantity is less than 0, receiving the current sampling, randomly overturning the spin in the spin configuration to update the spin configuration, and entering the next sampling round;
if the variable quantity is larger than 0, receiving the current sampling according to the probability of exp (-delta H/T), randomly overturning the spin in the spin configuration to update the spin configuration, and entering the next sampling round; wherein T is the current sampling temperature;
and when the last round of sampling is determined, enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result.
According to the data processing method based on the Itanium machine, when the difference value between the sampling temperature and 0K is smaller than a preset threshold value, the current sampling round is determined to be the last sampling round.
The invention also provides a data processing device based on the Itanium machine, which comprises:
the input module is used for circularly updating the spin configuration, modulating the updated spin configuration to the phase of the Gaussian beam and obtaining an input matrix; the spin configuration is generated randomly based on the constructed Esin model;
the output module is used for obtaining an output matrix according to the input matrix and the transformation matrix; wherein the transformation matrix is initially determined based on the constructed Exin model;
and the processing module is used for determining the output light intensity of the current sampling round according to the output matrix, determining the Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining a sampling result according to the Hamilton quantity, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
The invention also provides an Yixinji, comprising:
a wavefront modulator for:
circularly updating the spin configuration, and modulating the updated spin configuration to the phase of the Gaussian beam to obtain an input matrix, wherein the spin configuration is randomly generated based on the constructed Esino model;
separating the Gaussian beams and outputting the separated Gaussian beams;
coupling the separated light beams, and inputting the coupled Gaussian light beams into a light detector through a diaphragm and a lens;
the optical detector is used for obtaining an output matrix according to the input matrix and the transformation matrix; wherein the transformation matrix is initially determined based on the constructed Exin model;
and determining the output light intensity of the current sampling round according to the output matrix, determining a Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining a sampling result according to the Hamilton quantity, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
According to an embodiment of the present invention, the first wavefront modulator is specifically configured to:
modulating the updated spin configuration and the electrical feedback signal to the phase of the Gaussian beam, wherein the phase difference pi between the beam representing the +1 spin and the beam representing the-1 spin is the same, and the electric field amplitude of each beam is the same; the complex amplitudes of each beam are arranged to obtain an input matrix.
According to the present invention, there is provided an cooker, further comprising:
the first beam splitter is used for splitting laser to obtain reference light and object light, transmitting the object light to the first wavefront modulator and transmitting the reference light to the second beam splitter;
the second beam splitter is used for combining the reference light and the Gaussian beam coupled by the diaphragm and the lens and transmitting the combined light to the optical detector;
a first optical shutter for performing light field pattern acquisition on the reference light;
and the second optical shutter is used for carrying out light field pattern collection on the object light.
According to the data processing method and device based on the Italian machine and the Italian machine, provided by the invention, the data processing process of the Italian model can be completed on the light beam, the conversion from the optical signal to the electric signal can be realized, the capability of parallel processing of information at the light speed is realized, and the speed of solving the Italian problem can be greatly improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a data processing method based on an Itanium machine provided by the invention;
FIG. 2 is a schematic structural diagram of an Itanium-based data processing apparatus provided in the present invention;
FIG. 3 is a schematic structural diagram of an Yixinji machine provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present 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 data processing method and device based on the Italian machine and the Italian machine provided by the invention are described in the following with reference to FIGS. 1-3.
Fig. 1 shows a schematic flow chart of an ising machine-based data processing method provided by the present invention, and referring to fig. 1, the method includes:
11. circularly updating the spin configuration, and modulating the updated spin configuration to the phase of the Gaussian beam to obtain an input matrix; the spin configuration is generated randomly based on the constructed Esin model;
12. obtaining an output matrix according to the input matrix and the transformation matrix; wherein the transformation matrix is initially determined based on the constructed Exin model;
13. and determining the output light intensity of the current sampling round according to the output matrix, determining a Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining a sampling result according to the Hamilton quantity, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
With respect to step 11 to step 13, it should be noted that, in the present invention, the ixing model may be used to solve the combinatorial optimization problem. The combinatorial optimization problem refers to a class of problems that cannot find a globally optimal solution within polynomial time at present, such as a traveler problem and a maximum segmentation problem.
To this end, an Esinc model is constructed based on entity data for actual problems (e.g., traveler problem and majorit problem). And randomly generating an initial spin configuration based on the constructed Eschen model, modulating the initial spin configuration to the phase of a Gaussian beam of the Eschen machine, and realizing the purpose of encoding the spin of each point in the lattice of the Eschen model according to the phase of the beam.
And modulating the initial spin configuration to the phase of the Gaussian beam of the Itanium machine, and arranging the complex amplitude of the high-speed beam to obtain an input matrix of the Itanium machine on an input plane.
And then separating and coupling the Gaussian beams by an IshCi machine, wherein the process is equivalent to matrix transformation of an input matrix to obtain an output matrix of the IshCi machine on an output plane. The output matrix is obtained by calculating an input matrix and a transformation matrix, and the transformation matrix is also obtained based on the constructed Esino model configuration.
In the invention, the characteristic decomposition is carried out on the interaction coefficient matrix J of the Esinon model to obtain a transformation matrix A and a model characteristic value matrix I which are required in the separation and coupling process of Gaussian beamsλWherein J ═ A-1IλA。
In the invention, light beams output by an Esino machine are input into a light detector, the light detector determines the output light intensity of the current sampling round according to an output matrix, then determines a Hamilton value according to the output light intensity and a model characteristic value matrix, determines a sampling result according to the Hamilton value, the sampling result shows whether the initial spin configuration selected by the current sampling is proper or not, if the current sampling is received, the next sampling is carried out, the initial spin configuration is subjected to spin updating to obtain an updated spin configuration, then the updated spin configuration is executed again according to the processing process, then a sampling result is obtained, and then the sampling result is judged.
Therefore, the sampling round can be determined according to the sampling result, and the spinning configuration corresponding to the last round sampling is used as the data processing result when the last round sampling is determined.
In the invention, a spin configuration is randomly generated at the beginning, an input matrix is generated by an Eschen machine, the sampling Hamiltonian is obtained after the input matrix is subjected to space optical matrix transformation and electric domain calculation in each sampling, and the sampling is received according to the principle probability of a simulated annealing algorithm. At the end of sampling, it is possible to obtain a reasonable spin configuration as a result of data processing by the Esinon model to solve practical problems.
The data processing method based on the Itanium machine can complete the data processing process of the Itanium model on the light beam, can realize the conversion from the optical signal to the electric signal, has the capability of parallel processing of information at the light speed, and can greatly improve the speed of solving the Itanium problem.
In the further explanation of the above method, the process of modulating the updated spin configuration to the phase of the gaussian beam to obtain the input matrix is mainly explained as follows:
modulating the updated spin configuration and the electrical feedback signal to the phase of the Gaussian beam, wherein the phase difference pi between the beam representing the +1 spin and the beam representing the-1 spin is the same, and the electric field amplitude of each beam is the same; the complex amplitudes of each beam are arranged to obtain an input matrix.
In this regard, it should be noted that, in the present invention, the object beam is input to the wavefront modulator, and the wavefront modulator modulates the updated spin configuration and the electrical feedback signal onto the object beam to generate a plurality of gaussian beams, and each gaussian beam passes through the phase-encoded spin configuration on its beam. Wherein the phase difference between the beam characterizing the +1 spin and the beam characterizing the-1 spin is pi, and the electric field amplitude of each beam is the same. The complex amplitudes of each beam are then arranged to obtain an input matrix.
In the further explanation of the above method, the processing procedure of obtaining the output matrix according to the input matrix and the transformation matrix is mainly explained as follows:
and multiplying the input matrix and the transformation matrix to obtain an output matrix.
In this regard, it should be noted that, in the present invention, the iicin machine separates and couples gaussian beams, and the process is equivalent to matrix transformation of the input matrix. The input matrix is EinThen the wave front modulator splits and recombines the light beam, which is equivalent to matrix transformation A, and the output matrix on the output plane is marked as EoutThen there is Eout＝AEin。
In the further explanation of the above method, the processing procedure of determining the output light intensity of the current sampling round according to the output matrix and determining the hamilton quantity according to the output light intensity and the model eigenvalue matrix is mainly explained as follows:
determining the output light intensity of the current sampling round by adopting a first calculation formula according to the output matrix;
determining a Hamiltonian quantity by adopting a second calculation formula according to the output light intensity and the model characteristic value matrix;
wherein the first calculation formula includes:
I＝|AEin|2
where I is the output intensity, A is the transformation matrix, EinTo input a matrix, AEinIs an output matrix; l. capillary2Represents each of the pair matrixesTaking the square of an absolute value of an element;
the second calculation formula includes:
H(σ)＝-∑j(Iλ)jjIj
wherein σ represents the spin vector of the Esin model, H (σ) is the Hamiltonian, and IλIs a model eigenvalue matrix, (I)λ)jjIs IλJ-th diagonal element of (1)jIs the jth element of I.
In addition, if the second calculation formula needs to be corrected, a positive constant coefficient can be added to the formula on the right side of the formula, and the increase of the coefficient does not change the ground state spin configuration of the solved ircin problem.
Determining the variation between the Hamiltonian of the current sampling round and the Hamiltonian of the previous sampling round;
if the variable quantity is less than 0, receiving the current sampling, randomly overturning the spin in the spin configuration to update the spin configuration, and entering the next sampling round;
if the variable quantity is larger than 0, receiving the current sampling according to the probability of exp (-delta H/T), randomly overturning the spin in the spin configuration to update the spin configuration, and entering the next sampling round; and T is the current sampling temperature, the temperature T is not the real temperature and is only a virtual parameter, and the T is reduced according to a preset scheme after sampling is finished each time.
In the process of repeated sampling, the sampling temperature T needs to be slowly reduced according to a preset rule, and the sampling is terminated and the current spin configuration is output as the data processing result when the final T approaches to 0. Namely, when the difference between the sampling temperature and 0K (kelvin) is smaller than the preset threshold, the current sampling round is determined as the last sampling round.
The data processing device of the present invention is described below, and the data processing device of the present invention and the data processing method of the present invention can be referred to in correspondence with each other.
Fig. 2 shows a schematic structural diagram of an ising machine-based data processing apparatus provided by the present invention, referring to fig. 2, the apparatus includes an input module 21, an output module 22 and a processing module 23, wherein:
the input module 21 is configured to cyclically update the spin configuration, modulate the updated spin configuration to a phase of the gaussian beam, and obtain an input matrix; the spin configuration is generated randomly based on the constructed Esin model;
an output module 22, configured to obtain an output matrix according to the input matrix and the transformation matrix; wherein the transformation matrix is initially determined based on the constructed Exin model;
and the processing module 23 is configured to determine an output light intensity of the current sampling round according to the output matrix, determine a hamilton quantity according to the output light intensity and the model eigenvalue matrix, determine a sampling result according to the hamilton quantity, and enable a spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
In a further description of the above apparatus, the input module is specifically configured to, in a process of modulating the updated spin configuration to a phase of the gaussian beam to obtain an input matrix:
modulating the updated spin configuration and the electrical feedback signal to the phase of the Gaussian beam, wherein the phase difference pi between the beam representing the +1 spin and the beam representing the-1 spin is the same, and the electric field amplitude of each beam is the same;
the complex amplitudes of each beam are arranged to obtain an input matrix.
In a further description of the above apparatus, the output module, in a process of obtaining an output matrix according to the input matrix and the transformation matrix, is specifically configured to:
and multiplying the input matrix and the transformation matrix to obtain an output matrix.
In a further description of the above apparatus, the processing module is specifically configured to, during a processing procedure of determining an output light intensity of a current sampling round according to the output matrix and determining a hamilton quantity according to the output light intensity and the model eigenvalue matrix:
determining the output light intensity of the current sampling round by adopting a first calculation formula according to the output matrix;
determining a Hamiltonian quantity by adopting a second calculation formula according to the output light intensity and the model eigenvalue matrix;
wherein the first calculation formula includes:
I＝|AEin|2
where I is the output intensity, A is the transformation matrix, EinTo input a matrix, AEinIs an output matrix; l. capillary2Representing the square of the absolute value of each element in the matrix;
the second calculation formula includes:
H(σ)＝-∑j(Iλ)jjIj
wherein σ represents the spin vector of the Esin model, H (σ) is the Hamiltonian, and IλIs a model eigenvalue matrix, (I)λ)jjIs IλJ-th diagonal element of (1)jIs the jth element of I.
In a further description of the above apparatus, the processing module, when determining the sampling result according to the hamiltonian and when determining the last round of sampling, is specifically configured to:
determining the variation between the Hamiltonian of the current sampling round and the Hamiltonian of the previous sampling round;
if the variable quantity is less than 0, receiving the current sampling, randomly overturning the spin in the spin configuration to update the spin configuration, and entering the next sampling round;
if the variable quantity is larger than 0, receiving the current sampling according to the probability of exp (-delta H/T), randomly overturning the spin in the spin configuration to update the spin configuration, and entering the next sampling round; wherein T is the current sampling temperature;
and when the last round of sampling is determined, enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result.
In a further description of the above apparatus, when the difference between the sampling temperature and 0K is less than a preset threshold, the current sampling round is determined as the last sampling round.
Since the principle of the apparatus according to the embodiment of the present invention is the same as that of the method according to the above embodiment, further details are not described herein for further explanation.
It should be noted that, in the embodiment of the present invention, the relevant functional module may be implemented by a hardware processor (hardware processor).
The data processing device based on the Itanium machine can complete the data processing process of the Itanium model on the light beam, can realize the conversion from the optical signal to the electric signal, has the capability of parallel processing of information at the light speed, and can greatly improve the speed of solving the Itanium problem.
The invention provides an Yixing machine, which comprises the following structures:
a wavefront modulator for:
circularly updating the spin configuration, and modulating the updated spin configuration to the phase of the Gaussian beam to obtain an input matrix, wherein the spin configuration is randomly generated based on the constructed Esino model;
separating the Gaussian beams and outputting the separated Gaussian beams;
coupling the separated light beams, and inputting the coupled Gaussian light beams into a light detector through a diaphragm and a lens;
the optical detector is used for obtaining an output matrix according to the input matrix and the transformation matrix; the transformation matrix is generated randomly based on the constructed Exin model;
and determining the output light intensity of the current sampling round according to the output matrix, determining a Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining a sampling result according to the Hamilton quantity, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
In the present invention, the actions that can be performed by the wavefront modulator are referred to as action 1, action 2, action 3, and action 4 in this order.
Action 1: separating the Gaussian beams and outputting the separated Gaussian beams;
and action 2: re-separating the separated Gaussian beams, and outputting the re-separated Gaussian beams;
and action 3: coupling the separated light beams;
and 4, action: the spin configuration is modulated onto the phase of the gaussian beam.
In the present invention, three wavefront modulators are provided, named first wavefront modulator, second wavefront modulator and third wavefront modulator in this order. At this time, the preset execution mode includes:
the first wavefront modulator performs action 1, the second wavefront modulator performs action 2, the third wavefront modulator performs action 3, the first wavefront modulator performs action 4 or the second wavefront modulator performs action 4.
Alternatively, the first wavefront modulator performs act 1, the second wavefront modulator performs act 2, the third wavefront modulator performs act 3, the first wavefront modulator performs act 4 and the second wavefront modulator performs act 4 together
……
Here, all the cases of the preset execution manner are not explained.
The following is an explanation of one specific example, as shown in fig. 3, specifically as follows:
the first wavefront modulator 1 is configured to cyclically update a spin configuration, modulate the updated spin configuration on a phase of a gaussian beam to obtain an input matrix, separate the gaussian beam, and input the separated gaussian beam to the second wavefront modulator, where the spin configuration is initially determined based on a constructed isooctane model;
the second wave front modulator 2 is used for re-separating the separated Gaussian beams and inputting the re-separated Gaussian beams to the third wave front modulator;
the third wave front modulator 3 is used for coupling the separated light beams and inputting the coupled Gaussian light beams into the light detector through a diaphragm and a lens;
the optical detector 4 is used for obtaining an output matrix according to the input matrix and the transformation matrix; the transformation matrix is generated randomly based on the constructed Exin model;
and determining the output light intensity of the current sampling round according to the output matrix, determining a Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining a sampling result according to the Hamilton quantity, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
The first wavefront modulator is specifically configured to:
modulating the updated spin configuration and the electrical feedback signal to the phase of the Gaussian beam, wherein the phase difference pi between the beam representing the +1 spin and the beam representing the-1 spin is the same, and the electric field amplitude of each beam is the same; the complex amplitudes of each beam are arranged to obtain an input matrix.
The yi xin machine further comprises:
the first beam splitter 5 is configured to split laser light to obtain reference light and object light, transmit the object light to the first wavefront modulator, and transmit the reference light to the second beam splitter;
the second beam splitter 6 is used for combining the reference light and the Gaussian beam coupled by the diaphragm and the lens and transmitting the combined light to the optical detector;
a first optical shutter 7 for performing light field pattern acquisition on the reference light;
and the second optical shutter 8 is used for carrying out light field pattern collection on the object light.
In the invention, a monochromatic coherent light source 9 emits laser, the laser changes the linear polarization direction of input light through a half-wave plate, so that the polarization direction of the input light after passing through the half-wave plate is consistent with the polarization direction of a polarizer, and the polarization direction is consistent with the input polarization required by a wave front modulator. Then the light passes through the polarizer 10 and then is divided into the object light of the right path and the reference light of the left path by the first beam splitter. The object light is diffuse reflection light generated by irradiating the object with the laser light source, and includes information of the object. The reference light is light irradiated to the photosensitive film from the same laser light source through the half mirror. In each sampling process, the object light generates a group of Gaussian beams through the first wavefront modulator based on the electric feedback signal fed back by the optical detector and the spin configuration, and the spin configuration of the Eschen model is modulated on the phase of the beams. And then the object light is subjected to matrix transformation through a second spatial light modulator and a third wave front modulator to generate an output matrix, and the coupled Gaussian light beam is integrated with the reference light through a diaphragm 12, a lens 13, a second beam splitter and then sent to a light detector. Determining the output light intensity of the current sampling round by the optical detector according to the output matrix, determining the Hamilton quantity according to the output light intensity and the model characteristic value matrix, determining the sampling result according to the Hamilton quantity, determining the sampling round according to the sampling result, and enabling the spin configuration corresponding to the last round of sampling to be used as a data processing result when the last round of sampling is determined.
In the present invention, the delay line 11 in fig. 3 is used to adjust the optical path length of the reference light. The first and second optical shutters are used to measure the CCD pattern in the presence of object and reference light alone, and the complex amplitude of the output pattern, i.e., the output matrix, can be measured for calibration matrix transformations.
The Italic machine provided by the invention can complete the data processing process of the Italic model on a light beam, can realize the conversion from an optical signal to an electrical signal, has the capability of parallel processing of information at the light speed, and can greatly improve the speed of solving the Italic problem.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110797381.0A CN113722667B (en) | 2021-07-14 | 2021-07-14 | Data processing method and device based on Xin Ji and Xin Ji |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110797381.0A CN113722667B (en) | 2021-07-14 | 2021-07-14 | Data processing method and device based on Xin Ji and Xin Ji |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113722667A true CN113722667A (en) | 2021-11-30 |
| CN113722667B CN113722667B (en) | 2024-10-11 |
Family
ID=78673311
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110797381.0A Active CN113722667B (en) | 2021-07-14 | 2021-07-14 | Data processing method and device based on Xin Ji and Xin Ji |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113722667B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114489231A (en) * | 2022-02-02 | 2022-05-13 | 上海图灵智算量子科技有限公司 | Electric branching yixin machine capable of being programmed completely |
| CN114626536A (en) * | 2022-02-21 | 2022-06-14 | 华南理工大学 | A circuit for handling combinatorial optimization problems |
| CN115907005A (en) * | 2023-01-05 | 2023-04-04 | 华南理工大学 | A large-scale fully connected Ising model annealing circuit based on a network-on-chip |
| CN116341286A (en) * | 2023-05-24 | 2023-06-27 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | An FPGA-based accelerated quantum heuristic solution method and its device |
| WO2024037488A1 (en) * | 2022-08-16 | 2024-02-22 | 北京玻色量子科技有限公司 | Optical coherence computation apparatus and error correction method thereof |
| CN117993513A (en) * | 2024-02-07 | 2024-05-07 | 张江国家实验室 | Icine machine core circuit, unit thereof and method for annealing thereof |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2021060714A (en) * | 2019-10-04 | 2021-04-15 | 富士通株式会社 | Degree conversion device, degree conversion method, and degree conversion program |
| CN113093379A (en) * | 2021-03-25 | 2021-07-09 | 上海交通大学 | Photon Itanium machine-oriented orthogonal space phase modulation method |
-
2021
- 2021-07-14 CN CN202110797381.0A patent/CN113722667B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2021060714A (en) * | 2019-10-04 | 2021-04-15 | 富士通株式会社 | Degree conversion device, degree conversion method, and degree conversion program |
| CN113093379A (en) * | 2021-03-25 | 2021-07-09 | 上海交通大学 | Photon Itanium machine-oriented orthogonal space phase modulation method |
Non-Patent Citations (2)
| Title |
|---|
| D. PIERANGELI;G. MARCUCCI;C. CONTI: "Large-Scale Photonic Ising Machine by Spatial Light Modulation", vol. 122, no. 21, 31 December 2019 (2019-12-31) * |
| JIAYI OUYANG等: "On-demand photonic Ising machine with simplified Hamiltonian calculation by phase encoding and intensity detection", 《COMMUNICATIONS PHYSICS》, 24 May 2024 (2024-05-24) * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114489231A (en) * | 2022-02-02 | 2022-05-13 | 上海图灵智算量子科技有限公司 | Electric branching yixin machine capable of being programmed completely |
| CN114626536A (en) * | 2022-02-21 | 2022-06-14 | 华南理工大学 | A circuit for handling combinatorial optimization problems |
| WO2024037488A1 (en) * | 2022-08-16 | 2024-02-22 | 北京玻色量子科技有限公司 | Optical coherence computation apparatus and error correction method thereof |
| CN115907005A (en) * | 2023-01-05 | 2023-04-04 | 华南理工大学 | A large-scale fully connected Ising model annealing circuit based on a network-on-chip |
| CN116341286A (en) * | 2023-05-24 | 2023-06-27 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | An FPGA-based accelerated quantum heuristic solution method and its device |
| CN116341286B (en) * | 2023-05-24 | 2023-08-25 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Acceleration quantum heuristic solving method and device based on FPGA |
| CN117993513A (en) * | 2024-02-07 | 2024-05-07 | 张江国家实验室 | Icine machine core circuit, unit thereof and method for annealing thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113722667B (en) | 2024-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113722667A (en) | Data processing method and device based on Italian machine and Italian machine | |
| US12229662B2 (en) | All optical neural network | |
| US12026615B2 (en) | Training of photonic neural networks through in situ backpropagation | |
| CN111065963B (en) | Calculation device of Ito model | |
| Ikeda | Demonstration of quantum energy teleportation on superconducting quantum hardware | |
| EP4557169A1 (en) | Method and apparatus for optical system online designing based on intelligent light computing | |
| US11764828B2 (en) | Crosstalk reduction in multi-channel acousto-optic modulators | |
| Rambhatla et al. | Adaptive phase estimation through a genetic algorithm | |
| CN112639603A (en) | Spiking neural device and combined optimization problem calculation device | |
| Vischi et al. | Simulating photonic devices with noisy optical elements | |
| Anai et al. | Continuous-variable quantum kernel method on a programmable photonic quantum processor | |
| US20230185160A1 (en) | Super ising emulator with multi-body interactions and all-to-all connections | |
| US20230115065A1 (en) | Network depth-minimizing quantum compiler | |
| Nahar | Decoy-state quantum key distribution with arbitrary phase mixtures and phase correlations | |
| Wang et al. | Quantum next-generation reservoir computing and its quantum optical implementation | |
| Heo et al. | Procedure via cross-Kerr nonlinearities for encoding single logical qubit information onto four-photon decoherence-free states | |
| Zhang et al. | Mitigation of birefringence in cavity-based quantum networks using frequency-encoded photons | |
| Cui et al. | Transverse Polarization Gradient Entangling Gates for Trapped-Ion Quantum Computation | |
| Černotík et al. | Swap-test interferometry with biased qubit noise | |
| Gaur et al. | Information Processing in Hybrid Photonic Electrical Reservoir Computing | |
| CN120218262A (en) | Method, device, computer equipment and storage medium for pointer-free measurement of propagator | |
| Akel Abrahao | Frontiers of quantum optics: photonics tolls, computational complexity, quantum metrology, and quantum correlations | |
| Chen | Optical Quantum Computing | |
| Zhang et al. | Quadrature Photonic Spatial Ising Machine | |
| Laibacher | Multiphoton interference and the computational complexity of linear optics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |