CN113298246A - Data processing method, device and computer readable storage medium - Google Patents
Data processing method, device and computer readable storage medium Download PDFInfo
- Publication number
- CN113298246A CN113298246A CN202110583847.7A CN202110583847A CN113298246A CN 113298246 A CN113298246 A CN 113298246A CN 202110583847 A CN202110583847 A CN 202110583847A CN 113298246 A CN113298246 A CN 113298246A
- Authority
- CN
- China
- Prior art keywords
- optical signal
- data processing
- neural network
- processed
- determining
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/067—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
- G06N3/065—Analogue means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application discloses a data processing method, a data processing device and a computer readable storage medium. The method comprises the steps of determining a simulation mode of the nonlinear activation function in advance based on a data transmission mode of the nonlinear activation function of the photonic neural network and an optical path structure of the silicon-based chip, and determining optical path channels and phase shifter parameters corresponding to the photonic neural network based on the simulation mode. And inputting the acquired optical signal to be processed into a photonic neural network for optical signal transmission and optical signal calculation, and determining a data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network. The method and the device can effectively improve the calculation speed of the photon neural network, improve the data processing efficiency, reduce the power consumption, occupy less calculation resources and have low cost.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a data processing method and apparatus, and a computer-readable storage medium.
Background
With the development of Technology, society has now entered the era of cloud + AI (Artificial Intelligence) +5G (5th 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 manufacturing process of the semiconductor lithography is almost the physical limit of the moore's law at present, with the smaller and smaller manufacturing process, 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 problems of transmission bottleneck, power consumption increase, computing power bottleneck and the like of the existing von neumann and harvard architecture electronic computers are 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 a critical problem at present. 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.
The photonic operation of MZI (Mach-Zehnder interferometer) is the most common industrial solution in today's photonic neural networks, but it is only suitable for solving the multiply-add part based on 2 x 2 convolution operation and the extension based on 2 x 2 multiply-add operation. Although the largest amount of operations in an ANN (Artificial Neural Network) are derived from convolution operations, in order to ensure the accuracy of the operations, many nonlinear operations need to be added in a hidden layer, especially an activation function is heavy. Existing photonic computing networks have two solutions to activation function processing: 1. ONN (optical neural network) of MZI is used, but only convolution operation is carried out, all other parts need to complete photoelectric conversion, then silicon-based chips are used for realizing activation function operation, and then the activation function operation is converted into photons, and the photons return to the photonic neural network for carrying out other operations. Such a method requires a large amount of photoelectric conversion, is slow, consumes a large amount of power, and reduces the advantages of the photonic neural network. 2. Based on the characteristics of light energy, a special device is arranged to realize an activation function from the energy perspective. However, the special device is difficult to realize by the prior art, and has high cost and slow manufacturing process.
In view of this, how to solve the current situation that the speed of performing the data processing task by using the photonic neural network is slow and the occupied computational resources are large is a technical problem to be solved by those skilled in the art.
Disclosure of Invention
The application provides a data processing method, a data processing device and a computer readable storage medium, which can effectively improve the calculation speed of a photon neural network, improve the data processing efficiency, reduce the power consumption, occupy less calculation resources and have low cost.
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, including:
determining a simulation mode of a nonlinear activation function based on a data transmission mode of the nonlinear activation function of a photonic neural network and a light path structure of a silicon-based chip in advance, and determining a light path channel and phase shifter parameters corresponding to the photonic neural network based on the simulation mode;
acquiring an optical signal to be processed, and inputting the optical signal to be processed into the photonic neural network for calculation;
and determining a data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network.
Optionally, the determining a data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network includes:
determining a target light path channel of the photonic neural network according to the phase of the optical signal to be processed;
and obtaining a data processing result of the optical signal to be processed according to the output optical signal of the target optical path channel.
Optionally, the determining a target optical path channel of the photonic neural network according to the phase of the optical signal to be processed includes:
the optical path structure comprises an interferometer, wherein a first input end of the interferometer is switched on, and a second input end of the interferometer is switched off;
if the phase of the optical signal to be processed meets the single-channel gating condition, modulating the parameter of the phase shifter to be a preset value, and outputting the optical signal to be processed after phase modulation through the optical path channel corresponding to the first input end;
and if the phase of the optical signal to be processed does not meet the single-channel gating condition, calculating the optical signal to be processed according to a preset signal calculation relational expression.
Optionally, before determining the target optical path channel of the photonic neural network according to the phase of the optical signal to be processed, the method further includes:
and determining the single-channel gating condition according to the data transmission form of the nonlinear activation function.
Optionally, the determining, by the data transmission form of the nonlinear activation function based on the photonic neural network and the optical path structure of the silicon-based chip, the simulation mode of the nonlinear activation function includes:
determining a light conversion calculation relation of an input end and an output end based on the light path structure and the light conduction characteristics;
and determining a light path channel and a phase adjustment mode corresponding to the photonic neural network according to the light conversion calculation relation and the data transmission form of the nonlinear activation function.
Another aspect of an embodiment of the present invention provides a data processing apparatus, including:
the function simulation module is used for determining a simulation mode of the nonlinear activation function in advance based on a data transmission mode of the nonlinear activation function of the photonic neural network and a light path structure of the silicon-based chip so as to determine a light path channel and phase shifter parameters corresponding to the photonic neural network based on the simulation mode;
the optical signal processing module is used for inputting the acquired optical signal to be processed to the photonic neural network for calculation;
and the processing result determining module is used for determining the data processing result of the optical signal to be processed according to the output optical signal of the photon neural network.
Optionally, the processing result determining module is configured to: determining a target light path channel of the photonic neural network according to the phase of the optical signal to be processed; and obtaining a data processing result of the optical signal to be processed according to the output optical signal of the target optical path channel.
Optionally, the processing result determining module is further configured to: the optical path structure comprises an interferometer, wherein a first input end of the interferometer is switched on, and a second input end of the interferometer is switched off; if the phase of the optical signal to be processed meets the single-channel gating condition, modulating the parameter of the phase shifter to be a preset value, and outputting the optical signal to be processed after phase modulation through the optical path channel corresponding to the first input end; and if the phase of the optical signal to be processed does not meet the single-channel gating condition, calculating the optical signal to be processed according to a preset signal calculation relational expression.
An embodiment of the present invention further provides a data processing apparatus, which includes a processor, and the processor is configured to implement the steps of the data processing method according to any one of the foregoing when executing the computer program stored in the memory.
Finally, an embodiment of the present invention provides a computer-readable storage medium, where a data processing program is stored on the computer-readable storage medium, and when the data processing program is executed by a processor, the data processing method implements the steps of the data processing method according to any one of the foregoing items.
The technical scheme provided by the application has the advantages that the simulation scheme of the nonlinear activation function is determined through the light path structure corresponding to the photonic neural network and the data form of the nonlinear activation function adopted by the photonic neural network, and the corresponding light path channel and phase modulation angle are designed for the simulation part, so that the simulation of the characteristic of the nonlinear activation function is realized, the photonic neural network does not need to carry out a large amount of photoelectric conversion operation on the nonlinear activation function, the loss is effectively reduced, the occupied computing resource is correspondingly reduced, and the cost of the whole system is reduced. The speed and the performance of the photon computing network can be guaranteed without a large number of photoelectric conversion operations, and compared with the prior art, the speed of the photon neural network can be effectively increased, and the data processing efficiency is further improved.
In addition, the embodiment of the invention also provides a corresponding implementation device and a computer readable storage medium for the data processing method, so that the method has higher practicability, and the device and the computer readable storage medium 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 flow chart of a data processing method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of an optical path structure of an exemplary embodiment of the present invention;
fig. 3 is a schematic view of another optical path structure of an illustrative example provided by the embodiment of the present invention;
FIG. 4 is a diagram of a functional data form of an illustrative example provided by an embodiment of the present invention;
FIG. 5 is a schematic diagram of an optical path structure of a simulated nonlinear activation function according to an illustrative example provided by an embodiment of the present invention;
FIG. 6 is a block diagram of a data processing apparatus according to an embodiment of the present invention;
fig. 7 is a block diagram of another embodiment of a data processing apparatus 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. 1, fig. 1 is a schematic flow chart of a data processing method according to an embodiment of the present invention, where the embodiment of the present invention may include the following:
s101: and determining a simulation mode of the nonlinear activation function in advance based on a data transmission mode of the nonlinear activation function of the photonic neural network and the optical path structure of the silicon-based chip, and determining the optical path channel and phase shifter parameters corresponding to the photonic neural network based on the simulation mode.
In this step, the nonlinear activation function may be any nonlinear activation function in the neural network, such as a ReLU (Rectified Linear Unit), a Sigmoid function, a tanh function, and an empty ReLU (empty Rectified Linear Unit), where the data transmission form of the nonlinear activation function refers to a mathematical expression of the nonlinear activation function, and since the photonic neural network performs not only optical computation on the input optical signal, but also the most basic optical transmission task, the mathematical expression of the nonlinear activation function of the photonic neural network indicates that the photonic neural network is a data transmission form during the transmission process of the optical signal. The silicon-based chip is a photonic chip for realizing photonic computation, and integrates an electro-optical conversion device, an optical path system corresponding to a photonic neural network, a photoelectric conversion device, and the like, the electro-optical conversion device is a device for converting an electrical signal into an optical signal, such as a diode, and then the optical signal obtained by conversion is input into the optical path system, the optical path system is an optical path structure in the step, and for example, the optical path structure in the silicon-based chip based on MZI includes an MZI, a coupler, an attenuator, a phase shifter, and the like. The optical path system inputs an optical signal obtained through optical calculation into a photoelectric conversion device to obtain a corresponding electrical signal, and the photoelectric conversion device is a component for converting the optical signal into the electrical signal, such as a photoelectric detector. The simulation mode refers to what contents of the nonlinear activation function can be simulated, what contents cannot be simulated, and if the simulation can be carried out, corresponding channels are designed for the simulative parts and phase modulation is carried out through adjusting the phase shifter. The phase shifter parameters are the phase adjusting angles.
S102: and acquiring an optical signal to be processed, and inputting the optical signal to be processed into the photonic neural network for calculation.
The optical signal to be processed is data used for performing optical operation through a photonic neural network, the optical signal to be processed is generally obtained by conversion through an electro-optical conversion device, and optical signal data can also be directly obtained, the optical signal to be processed can be data in any form of image data, voice data and the like, and the image data can be remote sensing image data, microscopic imaging data and the like. In the process of data processing of the optical signal to be processed, the photon neural network directly performs nonlinear operation on the hidden layer of the optical signal to be processed, namely, the nonlinear activation function operation is directly performed in the simulation mode determined in the step S101, and the operation is directly performed in the photon neural network without performing secondary photoelectric conversion in the prior art or setting a special device.
S103: and determining a data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network.
It can be understood that, after the optical transmission and the optical operation are performed on the optical signal to be processed by the photonic neural network, a final calculation result is output through an output layer of the photonic neural network, and the calculation result is a final data processing result of the optical signal to be processed.
In the technical scheme provided by the embodiment of the invention, the simulation scheme of the nonlinear activation function is determined through the light path structure corresponding to the photonic neural network and the data form of the nonlinear activation function adopted by the photonic neural network, and the corresponding light path channel and phase modulation angle are designed for the simulatable part, so that the simulation of the characteristic of the nonlinear activation function is realized, the photonic neural network does not need to carry out a large number of photoelectric conversion operations on the nonlinear activation function, the loss is effectively reduced, the occupied computing resources are correspondingly reduced, and the cost of the whole system is reduced. The speed and the performance of the photon computing network can be guaranteed without a large number of photoelectric conversion operations, and compared with the prior art, the speed of the photon neural network can be effectively increased, and the data processing efficiency is further 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. 1 is only an exemplary manner, and does not represent that only the execution order is the order.
In the foregoing embodiment, how to execute the step of processing the optical signal by the photonic neural network is not limited, and an implementation manner is provided in this embodiment, and may include the following steps:
in this step, the analog mode of the nonlinear activation function is related to the phase of the optical signal, so that after the optical signal is input to the photonic neural network, the current phase of the optical signal to be processed is obtained first, and the target optical path channel of the photonic neural network is determined according to the phase of the optical signal to be processed; and obtaining a data processing result of the optical signal to be processed according to the output optical signal of the target optical path channel.
In this embodiment, the optical path structure includes an interferometer and a phase shifter, where two optical signals of the interferometer are phase-modulated by the phase shifter, and finally data is output at an output end. That is, the interferometer includes two optical paths, and in order to distinguish the first and second optical paths for two pipelines, the first optical path is an optical path that is output from the first output terminal after the phase modulation of the input optical signal of the first input terminal, and the second optical path is an optical path that is output from the second output terminal after the phase modulation of the input optical signal of the second input terminal. In this embodiment, a determination condition of which optical path to go is determined based on the data transmission form of the nonlinear activation function and the phase of the optical signal, as an optional implementation manner, a single-channel gating condition may be determined according to the data transmission form of the nonlinear activation function, where the single-channel gating condition is determined based on the phase of the optical signal. After an optical signal to be processed enters a photonic neural network, phase information of the optical signal to be processed is obtained, if the phase of the optical signal to be processed meets a single-channel gating condition, a modulation phase shifter parameter is a preset value, and the optical signal to be processed is output after being subjected to phase modulation through a light path channel corresponding to a first input end; and if the phase of the optical signal to be processed does not meet the single-channel gating condition, calculating the optical signal to be processed according to a preset signal calculation relational expression. The predetermined signal calculation relationship is also predetermined based on the data transmission form of the nonlinear activation function and the optical signal phase.
In the foregoing embodiment, the simulation manner of the nonlinear activation function of the photonic neural network is not limited, and an implementation manner is provided in this embodiment, and may include the following steps:
determining a light conversion calculation relation between an input end and an output end based on the light path structure and the light conduction characteristic; and determining a light path channel corresponding to the photonic neural network and a phase adjustment mode according to the light conversion calculation relation and the data transmission form of the nonlinear activation function.
The optical transmission characteristic means that in the transmission process of the photonic neural network, the real part and the imaginary part are required to be involved in transmission by combining the mathematical expression form of an optical signal, each operation is that the imaginary part is required to participate as energy loss, and in the photoelectric conversion output, only the real part of the light can be identified. The optical conversion calculation relationship refers to the processing operation of the optical signal involved in the process that the optical signal is input to the output end through the input end and output.
In order to make the technical solutions of the present application more clearly apparent to those skilled in the art, in the present application, a nonlinear activation function is taken as a Relu function, and a photonic neural network is illustrated by taking an example of a photonic operation based on a MZI interferometer as an example, in this embodiment, an optical path structure includes an MZI interferometer, a phase shifter, a coupler and an attenuator, two optical signals of the MZI interferometer are both phase-modulated by the phase shifter and finally output data at an output end, the MZI interferometer includes two optical path channels, in order to distinguish between a first optical path channel and a second optical path channel for two pipelines, the first optical path channel is an optical path in which an input optical signal at a first input end is phase-modulated and then output through a first output end, and the second optical path channel is an optical path in which an input optical signal at a second input end is phase-modulated and then output through a second output.
The forward propagation process of the artificial neural network strongly depends on multiply-add operation, and most of the operation in the inference process is essentially linear operation between the trained weight and the characteristic value. The use of optical chips to compute matrix multiplication is very different from electrical chips in terms of implementation principles. In digital integrated circuits, data is typically encoded as binary strings in the switching states of transistors. The numbers represented by binary strings are discrete, e.g., integers or floating point values; in photonics, data is encoded by modulating the amplitude or phase of a laser pulse, resulting in a continuous real value, changing the intensity or phase of the optical field changes the real number represented. The circuit can use conducting wire to guide electron, and the photonics can use silicon-based optical waveguide structure to transmit laser. On a mathematical model, a programmable phase shifter, a mach-zehnder interferometer, or other structures can implement matrix multiplication of any dimension in the optical domain by using a characteristic Decomposition method such as SVD (Singular Value Decomposition). In linear algebra, singular value decomposition is an important matrix decomposition mode, is one of algorithms commonly used in machine learning, and is widely applied to feature extraction, data simplification and recommendation systems. The real number matrix of any dimensionality can be decomposed into the product of three matrixes through a singular value decomposition method. Assuming that M is a matrix of M × n, U is a matrix of M × M, called unitary matrix, Σ is a diagonal matrix of M × n, values on the diagonal are non-negative real numbers, V is a matrix of n × n, also called unitary matrix, the complex conjugate matrix of V is represented by V, and the singular value decomposition of matrix M can be represented by formula (1):
M＝U∑V*。 (1)
based on the above theoretical basis, for the structure shown in fig. 2, L1 and L2 are the optical inputs of the MZI, and the output optical signals are L1 'and L2'. The MZI can couple the optical power of one dual port to the optical power of another dual port in a certain proportion, the splitting ratio is 50: 50. 2 θ is a phase shifter, which has a programmable function, and can be implemented in several ways, such as plating a metal film on a section of waveguide material, and applying an external voltage to control a metal film heater to cause a waveguide temperature change to change a refractive index, thereby implementing a phase shift; phase shift can also be introduced by altering the waveguide refractive index using the plasma dispersion effect, i.e., changing the concentration of electrons and holes and the electro-optic effect. If A represents amplitude, ω represents frequency, t represents time, and θ represents initial phase, L1 and L2 are expressed as
In the formula, A1、A2、θ1And theta2The sum of the amplitudes of the optical signal L1 of the first optical path and the optical signal L2 of the second optical pathThe initial phase. Since only the real part of light can be identified during the photoelectric conversion process, and the imaginary part represents the energy loss during transmission, equation (2) above can be converted based on the euler equation:
in order to develop the above formula to obtain the transmission relationship, parameters S1, S2, S3 and S4 may be set for the transmission intermediate state of fig. 2, and the relationship is shown in fig. 3. S1 is an optical signal of the first optical path before entering the phase shifter, S2 is an optical signal of the second optical path before entering the phase shifter, S3 is an optical signal of the first optical path after being output from the phase shifter, and S4 is an optical signal of the second optical path after being output from the phase shifter. The expression (3) Re represents the real part. After L1 and L2 enter the MZI, it is known that the energy contained in the light is transferred to the optical signals corresponding to the two output ports through the coupler, and thus the amplitudes thereof are the original ones
The relationships of L1 and L2 to S1 and S2 in fig. 3 can thus be represented by equation (4):
since the transmission needs to involve the real part and the imaginary part for the light conduction, the imaginary part needs to be involved as energy loss in each operation, and only the real part needs to be considered for the output of the photoelectric conversion, the arrangement is such that
To meet the requirements of actual operation. Based on the coupler relationship, it can know
And [ S1, S2 ]]TThe corresponding relation is as follows:
where j is an imaginary number, c denotes that the imaginary part is not taken into account, T denotes a matrix transpose,
to represent complex operations that do not consider imaginary parts for L1, L2. Based on the operation mode of the attenuator with MZI and the structure shown in FIG. 3, the expressions S3 and S4 can be obtained as follows:
based on the above, the calculation relationship between L1 and L2 and the corresponding L1 'and L2' under such operation, that is, the calculation relationship between the light conversion at the input end and the light conversion at the output end, can be finally obtained:
after determining the optical conversion calculation relationship between the input end and the output end of the optical path structure corresponding to the photonic neural network, the implementation process of determining the optical path channel corresponding to the photonic neural network and the phase adjustment mode according to the optical conversion calculation relationship and the data transmission form of the nonlinear activation function may include the following steps:
in this embodiment, taking an activation function Relu as an example, as shown in fig. 4, a data transmission format of the activation function Relu takes a value of 0 when x is less than 0, and y ═ f (x) when x > 0. To determine the simulation mode of the activation function Relu, the input-output relationship summarized in the above equation (7) may be first expanded to obtain:
since the light transmission is involved here and no photoelectric conversion works, the calculation needs to be done in a full mode containing the real and imaginary parts. Based on the real part and imaginary part relation, the operation of L1 and L2 based on euler's formula and formula (3) can be deduced:
in the formula (I), the compound is shown in the specification,
to the phase of the optical signal entering the first optical path, θ is a phase modulation value of the phase shifter. The second half of the equation (8) relates to the case where the L2 part is 0, that is, the MZI at this time is operated as a single optical path, so equation (9) can obtain the phase-modulated output of L1' to L1 at the MZI. The data information that can be obtained as photoelectric conversion in the real part can be obtained as:
therefore, when the phase modulation value 2 θ of the MZI phase shifter in fig. 3 is 0, all light transmission is transmitted only through the L1 path, and the output relationship can be obtained that satisfies the output requirement that the left side y of the activation function Relu is 0. Further considering the observation information obtained by photoelectric conversion of L1 by the euler formula is:
amplitude A1The number of photons is shown, taking the square root of the light intensity, which must be positive. The phase of light is 0-2 pi, so that to realize the activation function Relu, the operation mode based on MZI can be expressed as:
an analogous way of implementing the activation function Relu is obtained on the basis of equation 12 above,in the simulation mode, the optical path structure of the MZI is as shown in fig. 5, the L2 end of the MZI, i.e. the second optical path channel, encloses possible interference light, L1' is used as output, after the optical signal to be processed is input into the optical path structure, the phase determination is performed first, and when the phase of the optical signal to be processed satisfies the requirement
And in the time, L1' gates the output of the MZI, namely the MZI is a single-channel optical path, and the optical signal to be processed is phase-modulated through a first optical path and then is output through a first output end. When the phase of the optical signal to be processed does not satisfy
Then, the formula of the system is used
Calculating the optical signal to be processed, and outputting the calculated result as the final result through the judged channel, wherein A is the amplitude of the optical signal to be processed,
is the phase of the optical signal to be processed. In this embodiment, the predetermined signal calculation relation can be expressed as
As can be seen from the above, the artificial neural network scheme implemented by photon operation of MZI is used as a basis in this embodiment, the conditions that Relu can be implemented after conversion by MZI are obtained by combining matrix operation formula and characteristic derivation of optical operation and euler expansion and calculation of light, so as to implement a simulation solution applicable to nonlinear activation function Relu in the photonic neural network, and based on the optical conversion calculation relationship of MZI and the L1 and L2 setting relationship of this relationship, the phase is adjusted, and based on the implementation manner of phase judgment, the scheme of activation function Relu can be replaced without performing a large number of photoelectric conversions, so as to ensure the speed and performance of the photonic calculation network, and have the advantages of rapidness and low power consumption.
The embodiment of the invention also provides a corresponding device for the data processing method, thereby further ensuring 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. In the following, the data processing apparatus provided by the embodiment of the present invention is introduced, and the data processing apparatus described below and the data processing method described above may be referred to correspondingly.
Based on the angle of the functional module, referring to fig. 6, fig. 6 is a structural diagram of a data processing apparatus according to an embodiment of the present invention, in a specific implementation manner, the apparatus may include:
the function simulation module 601 is configured to determine a simulation mode of the nonlinear activation function in advance based on a data transmission form of the nonlinear activation function of the photonic neural network and an optical path structure of the silicon-based chip, so as to determine optical path channels and phase shifter parameters corresponding to the photonic neural network based on the simulation mode.
And the optical signal processing module 602 is configured to input the acquired optical signal to be processed to the photonic neural network for calculation.
A processing result determining module 603, configured to determine a data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network.
Optionally, in some embodiments of this embodiment, the processing result determining module 603 may be configured to: determining a target light path channel of the photonic neural network according to the phase of the optical signal to be processed; and obtaining a data processing result of the optical signal to be processed according to the output optical signal of the target optical path channel.
As an optional implementation manner of this embodiment, the processing result determining module may be further configured to: the nonlinear activation function is a Relu function, the optical path structure comprises an interferometer, a first input end of the interferometer is connected, and a second input end of the interferometer is disconnected; if the phase of the optical signal to be processed meets the single-channel gating condition, the parameter of the modulation phase shifter is a preset value, and the optical signal to be processed is output after being phase-modulated through a light path channel corresponding to the first input end; and if the phase of the optical signal to be processed does not meet the single-channel gating condition, calculating the optical signal to be processed according to a preset signal calculation relational expression.
As another optional implementation manner of this embodiment, the processing result determining module 603 may further include a condition determining unit, configured to determine the single-channel gating condition according to a data transmission form of the nonlinear activation function.
Optionally, in other embodiments of this embodiment, the function simulation module 601 may be further configured to: determining a light conversion calculation relation between an input end and an output end based on the light path structure and the light conduction characteristic; and determining a light path channel corresponding to the photonic neural network and a phase adjustment mode according to the light conversion calculation relation and the data transmission form of the nonlinear activation function.
The functions of the functional modules of the data processing apparatus according to the embodiment of the present invention may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the description related to the foregoing method embodiment, which is not described herein again.
Therefore, the embodiment of the invention not only can effectively improve the calculation speed of the photon neural network and improve the data processing efficiency, but also can reduce the power consumption, occupy less calculation resources and have low cost.
The data processing device mentioned above is described from the perspective of functional modules, and further, the present application also provides a data processing device described from the perspective of hardware. Fig. 7 is a block diagram of another data processing apparatus according to an embodiment of the present application. As shown in fig. 7, the apparatus comprises a memory 70 for storing a computer program; a processor 71, configured to implement the steps of the data processing method according to any of the above embodiments when executing the computer program.
The processor 71 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like, among others. The processor 71 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 71 may also include a main processor and a coprocessor, 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 71 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 71 may further include an AI (Artificial Intelligence) processor for processing computing operations related to machine learning.
The memory 70 may include one or more computer-readable storage media, which may be non-transitory. Memory 70 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 70 is at least used for storing the following computer program 701, wherein after being loaded and executed by the processor 71, the computer program can implement the relevant steps of the data processing method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 70 may also include an operating system 702, data 703, and the like, and the storage manner may be a transient storage or a permanent storage. Operating system 702 may include Windows, Unix, Linux, etc. The data 703 may include, but is not limited to, data corresponding to a result of data processing, and the like.
In some embodiments, the data processing device may further include a display 72, an input/output interface 73, a communication interface 74, otherwise known as a network interface, a power supply 75, and a communication bus 76. The display 72 and the input/output interface 73, 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 data processing device and for displaying a visual user interface. The communication interface 74 may optionally include a wired interface and/or a wireless interface, such as a WI-FI interface, a bluetooth interface, etc., typically used to establish a communication link between the data processing apparatus and other electronic devices. The communication bus 76 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. 7, but this is not intended to represent only one bus or type of bus.
Those skilled in the art will appreciate that the configuration shown in fig. 7 is not intended to be limiting of the data processing apparatus and may include more or fewer components than those shown, such as a sensor 77 that performs various functions.
The functions of the functional modules of the data processing apparatus according to the embodiment of the present invention may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the description related to the foregoing method embodiment, which is not described herein again.
Therefore, the embodiment of the invention not only can effectively improve the calculation speed of the photon neural network and improve the data processing efficiency, but also can reduce the power consumption, occupy less calculation resources and have low cost.
It is to be understood that, if the data processing method in the above-described 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, and the data processing program is executed by a processor, and the steps of the data processing method according to any one of the above embodiments are provided.
The functions of the functional modules of the computer-readable storage medium according to the embodiment of the present invention 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 not only can effectively improve the calculation speed of the photon neural network and improve the data processing efficiency, but also can reduce the power consumption, occupy less calculation resources and have low cost.
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.
A data processing method, an apparatus and a computer-readable storage medium 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.
Claims (10)
1. A data processing method, comprising:
determining a simulation mode of a nonlinear activation function based on a data transmission mode of the nonlinear activation function of a photonic neural network and a light path structure of a silicon-based chip in advance, and determining a light path channel and phase shifter parameters corresponding to the photonic neural network based on the simulation mode;
acquiring an optical signal to be processed, and inputting the optical signal to be processed into the photonic neural network for calculation;
and determining a data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network.
2. The data processing method of claim 1, wherein the determining the data processing result of the optical signal to be processed according to the output optical signal of the photonic neural network comprises:
determining a target light path channel of the photonic neural network according to the phase of the optical signal to be processed;
and obtaining a data processing result of the optical signal to be processed according to the output optical signal of the target optical path channel.
3. The data processing method of claim 2, wherein the determining a target optical path channel of the photonic neural network according to the phase of the optical signal to be processed comprises:
the optical path structure comprises an interferometer, wherein a first input end of the interferometer is switched on, and a second input end of the interferometer is switched off;
if the phase of the optical signal to be processed meets the single-channel gating condition, modulating the parameter of the phase shifter to be a preset value, and outputting the optical signal to be processed after phase modulation through the optical path channel corresponding to the first input end;
and if the phase of the optical signal to be processed does not meet the single-channel gating condition, calculating the optical signal to be processed according to a preset signal calculation relational expression.
4. The data processing method of claim 3, wherein before determining the target optical path channel of the photonic neural network according to the phase of the optical signal to be processed, the method further comprises:
and determining the single-channel gating condition according to the data transmission form of the nonlinear activation function.
5. The data processing method according to any one of claims 1 to 4, wherein the determining the simulation mode of the nonlinear activation function based on the data transmission form of the nonlinear activation function of the photonic neural network and the optical path structure of the silicon-based chip comprises:
determining a light conversion calculation relation of an input end and an output end based on the light path structure and the light conduction characteristics;
and determining a light path channel and a phase adjustment mode corresponding to the photonic neural network according to the light conversion calculation relation and the data transmission form of the nonlinear activation function.
6. A data processing apparatus, comprising:
the function simulation module is used for determining a simulation mode of the nonlinear activation function in advance based on a data transmission mode of the nonlinear activation function of the photonic neural network and a light path structure of the silicon-based chip so as to determine a light path channel and phase shifter parameters corresponding to the photonic neural network based on the simulation mode;
the optical signal processing module is used for inputting the acquired optical signal to be processed to the photonic neural network for calculation;
and the processing result determining module is used for determining the data processing result of the optical signal to be processed according to the output optical signal of the photon neural network.
7. The data processing apparatus of claim 6, wherein the processing result determination module is configured to: determining a target light path channel of the photonic neural network according to the phase of the optical signal to be processed; and obtaining a data processing result of the optical signal to be processed according to the output optical signal of the target optical path channel.
8. The data processing apparatus of claim 7, wherein the processing result determination module is further configured to: the optical path structure comprises an interferometer, wherein a first input end of the interferometer is switched on, and a second input end of the interferometer is switched off; if the phase of the optical signal to be processed meets the single-channel gating condition, modulating the parameter of the phase shifter to be a preset value, and outputting the optical signal to be processed after phase modulation through the optical path channel corresponding to the first input end; and if the phase of the optical signal to be processed does not meet the single-channel gating condition, calculating the optical signal to be processed according to a preset signal calculation relational expression.
9. A data processing apparatus comprising a processor for implementing the steps of the data processing method according to any one of claims 1 to 5 when executing a computer program stored in a memory.
10. A computer-readable storage medium, on which a data processing program is stored, which when executed by a processor implements the steps of the data processing method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110583847.7A CN113298246B (en) | 2021-05-27 | 2021-05-27 | Data processing method, device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110583847.7A CN113298246B (en) | 2021-05-27 | 2021-05-27 | Data processing method, device and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113298246A true CN113298246A (en) | 2021-08-24 |
| CN113298246B CN113298246B (en) | 2023-02-28 |
Family
ID=77325431
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110583847.7A Active CN113298246B (en) | 2021-05-27 | 2021-05-27 | Data processing method, device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113298246B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822425A (en) * | 2021-11-02 | 2021-12-21 | 广西大学 | Integrated optical neural network system capable of resisting transmission loss interference |
| CN114037070A (en) * | 2022-01-07 | 2022-02-11 | 苏州浪潮智能科技有限公司 | Optical signal processing method, photonic neural network chip and design method thereof |
| CN115049050A (en) * | 2022-08-11 | 2022-09-13 | 北京冠邦开源智能系统工程技术有限公司 | Artificial intelligence oriented neural unit implementation method and circuit |
| WO2023109065A1 (en) * | 2021-12-15 | 2023-06-22 | 苏州浪潮智能科技有限公司 | Method and system for implementing recurrent neural network, and electronic device and storage medium |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109272110A (en) * | 2018-10-31 | 2019-01-25 | 上海交通大学 | Photoelectricity based on photon neural network chip merges intelligent signal processing system |
| CN109314683A (en) * | 2016-08-29 | 2019-02-05 | Ntt 电子株式会社 | Optical transport device for compensation of distortion, optical transport distortion compensating method and communication device |
| CN109477938A (en) * | 2016-06-02 | 2019-03-15 | 麻省理工学院 | Device and method for optical neural network |
| CN109871871A (en) * | 2019-01-16 | 2019-06-11 | 南方科技大学 | Image-recognizing method, device and electronic equipment based on optical neural network structure |
| CN110197277A (en) * | 2019-05-13 | 2019-09-03 | 浙江大学 | Realize the optical neural network method of number identification |
| CN110503196A (en) * | 2019-08-26 | 2019-11-26 | 光子算数(北京)科技有限责任公司 | A kind of photon neural network chip and data processing system |
| CN110728358A (en) * | 2019-09-30 | 2020-01-24 | 上海商汤智能科技有限公司 | Data processing method and device based on neural network |
| CN110956259A (en) * | 2019-11-22 | 2020-04-03 | 联合微电子中心有限责任公司 | Photon neural network training method based on forward propagation |
| CN111680796A (en) * | 2020-06-29 | 2020-09-18 | 中国人民解放军国防科技大学 | On-chip photonic neural network device based on cascade optical coupler, chip and application method thereof |
| CN111860822A (en) * | 2020-07-20 | 2020-10-30 | 联合微电子中心有限责任公司 | All-optical nonlinear activation function implementation method and device of optical neural network |
| CN112232504A (en) * | 2020-09-11 | 2021-01-15 | 联合微电子中心有限责任公司 | Photon neural network |
| CN112529166A (en) * | 2020-12-25 | 2021-03-19 | 中国科学院西安光学精密机械研究所 | Fusion neuron model, neural network structure, training and reasoning method, storage medium and device |
-
2021
- 2021-05-27 CN CN202110583847.7A patent/CN113298246B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109477938A (en) * | 2016-06-02 | 2019-03-15 | 麻省理工学院 | Device and method for optical neural network |
| CN109314683A (en) * | 2016-08-29 | 2019-02-05 | Ntt 电子株式会社 | Optical transport device for compensation of distortion, optical transport distortion compensating method and communication device |
| CN109272110A (en) * | 2018-10-31 | 2019-01-25 | 上海交通大学 | Photoelectricity based on photon neural network chip merges intelligent signal processing system |
| CN109871871A (en) * | 2019-01-16 | 2019-06-11 | 南方科技大学 | Image-recognizing method, device and electronic equipment based on optical neural network structure |
| CN110197277A (en) * | 2019-05-13 | 2019-09-03 | 浙江大学 | Realize the optical neural network method of number identification |
| CN110503196A (en) * | 2019-08-26 | 2019-11-26 | 光子算数(北京)科技有限责任公司 | A kind of photon neural network chip and data processing system |
| CN110728358A (en) * | 2019-09-30 | 2020-01-24 | 上海商汤智能科技有限公司 | Data processing method and device based on neural network |
| CN110956259A (en) * | 2019-11-22 | 2020-04-03 | 联合微电子中心有限责任公司 | Photon neural network training method based on forward propagation |
| CN111680796A (en) * | 2020-06-29 | 2020-09-18 | 中国人民解放军国防科技大学 | On-chip photonic neural network device based on cascade optical coupler, chip and application method thereof |
| CN111860822A (en) * | 2020-07-20 | 2020-10-30 | 联合微电子中心有限责任公司 | All-optical nonlinear activation function implementation method and device of optical neural network |
| CN112232504A (en) * | 2020-09-11 | 2021-01-15 | 联合微电子中心有限责任公司 | Photon neural network |
| CN112529166A (en) * | 2020-12-25 | 2021-03-19 | 中国科学院西安光学精密机械研究所 | Fusion neuron model, neural network structure, training and reasoning method, storage medium and device |
Non-Patent Citations (3)
| Title |
|---|
| B. J. SHASTRI等: "Silicon Photonic Neural Networks and Applications", 《2020 PHOTONICS NORTH (PN)》 * |
| X. XIAO ， S. J. BEN YOO: "Tensor-Train Decomposed Synaptic Interconnections for Compact and Scalable Photonic Neural Networks", 《2020 IEEE PHOTONICS CONFERENCE (IPC)》 * |
| 陈媛: "基于硅基光波导实现特征识别的光子神经网络仿真研究", 《中国优秀博硕士学位论文全文数据库(硕士) 基础科学辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822425A (en) * | 2021-11-02 | 2021-12-21 | 广西大学 | Integrated optical neural network system capable of resisting transmission loss interference |
| CN113822425B (en) * | 2021-11-02 | 2023-07-18 | 广西大学 | Transmission loss interference resistant integrated optical neural network system |
| WO2023109065A1 (en) * | 2021-12-15 | 2023-06-22 | 苏州浪潮智能科技有限公司 | Method and system for implementing recurrent neural network, and electronic device and storage medium |
| CN114037070A (en) * | 2022-01-07 | 2022-02-11 | 苏州浪潮智能科技有限公司 | Optical signal processing method, photonic neural network chip and design method thereof |
| CN115049050A (en) * | 2022-08-11 | 2022-09-13 | 北京冠邦开源智能系统工程技术有限公司 | Artificial intelligence oriented neural unit implementation method and circuit |
| CN115049050B (en) * | 2022-08-11 | 2022-11-22 | 北京冠邦开源智能系统工程技术有限公司 | Artificial intelligence oriented neural unit implementation method and circuit |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113298246B (en) | 2023-02-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113298246B (en) | Data processing method, device and computer readable storage medium | |
| Liu et al. | Holylight: A nanophotonic accelerator for deep learning in data centers | |
| Qu et al. | Inverse design of an integrated-nanophotonics optical neural network | |
| CN113627605A (en) | Optical interference unit configuration method and device of photonic neural network and storage medium | |
| US20160247062A1 (en) | Mapping of algorithms to neurosynaptic hardware | |
| CN113408720B (en) | Optical neural network, data processing method and device based on optical neural network, and storage medium | |
| WO2023005084A1 (en) | Optical circuit building method, optical circuit, and optical signal processing method and apparatus | |
| Cheng et al. | Logarithm-approximate floating-point multiplier is applicable to power-efficient neural network training | |
| Chakravarty et al. | Numerical studies of the KP line-solitons | |
| Moralis-Pegios et al. | Photonic neuromorphic computing: Architectures, technologies, and training models | |
| Giamougiannis et al. | Neuromorphic silicon photonics with 50 GHz tiled matrix multiplication for deep-learning applications | |
| US20220043474A1 (en) | Path-number-balanced universal photonic network | |
| Huang et al. | Easily scalable photonic tensor core based on tunable units with single internal phase shifters | |
| Li et al. | Research and implementation of parallel artificial fish swarm algorithm based on ternary optical computer | |
| Kai et al. | The symmetric MSD encoder for one-step adder of ternary optical computer | |
| CN113592084B (en) | Photon neural network on chip based on reverse optimization superstructure convolution kernel | |
| CN113325650B (en) | Optical circuit, optical signal processing method, optical signal processing device and readable storage medium | |
| CN113392966B (en) | Method, equipment and storage medium for realizing neural network average pooling | |
| CN115905792A (en) | Mach-Zehnder interferometer network for optical real number matrix calculation | |
| US20190325289A1 (en) | Optimizing performance of recurrent neural networks | |
| Hirose et al. | Complex-valued neural networks for wave-based realization of reservoir computing | |
| CN107066423A (en) | A kind of optimization method of limited input fixed-point number matrix multiplier | |
| Chen et al. | Optical logic operations using a meshed Mach-Zehnder interference array | |
| Hattori et al. | Optical-electronic implementation of artificial neural network for ultrafast and accurate inference processing | |
| Shah et al. | AnalogVNN: A fully modular framework for modeling and optimizing photonic neural networks |
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 |