Data Processing Method and Device Based on Pulse Rearrangement Deep Residual Neural Network

Abstract

This application relates to the technical field of spiking neural networks and data processing. More specifically, this application relates to a data processing method and apparatus based on a pulse rearrangement deep residual neural network. The method includes: configuring a plurality of adjacent pulse rearrangement residual modules in a pulse rearrangement deep residual neural network; using the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as a first pulse rearrangement model; training the first pulse rearrangement model; inputting target data into the trained first pulse rearrangement model for processing to obtain a target processing result. The rearrangement processing in this application greatly reduces the number of parameters, reduces the risk of overfitting, and also reduces the storage and calculation overhead, thereby improving the efficiency of data processing. At the same time, this application can obtain the category corresponding to the target data, and can obtain a regression sequence or a single regression vector, so as to be applicable to various application scenarios.

Description

Technical Field

[0001] This application relates to the technical fields of spiking neural networks and data processing. More specifically, this application relates to a data processing method and apparatus based on a spiking rearrangement deep residual neural network. Background Art

[0002] Artificial Neural Networks (ANNs) have made breakthroughs in many fields, thanks to deep learning. The depth of the network has a great impact on the performance of the network, and deeper network structures have been successively proposed and have better performance than shallow networks. In deep ANNs, the most common structure is residual connections, which successfully solves the training problem of deep ANNs.

[0003] Spiking Neural Networks (SNNs) are known as the third generation of neural networks and have the advantages of event-driven and low power consumption. However, due to the use of discrete spikes for communication in SNNs, there is a large amount of information loss. To improve the performance of SNNs, a natural idea is to use a structure similar to the residual network to obtain performance gains by increasing the depth. However, directly using the residual network structure in traditional ANNs still results in poor performance of SNNs. Summary of the Invention

[0004] Based on the above technical problems, the present invention aims to configure a plurality of adjacent spiking rearrangement residual modules based on a spiking rearrangement deep residual neural network, use the spiking rearrangement deep residual neural network after configuring the spiking rearrangement residual blocks as a first spiking rearrangement model, and process target data using the first spiking rearrangement model.

[0005] The first aspect of the present invention provides a data processing method based on a spiking rearrangement deep residual neural network, the method comprising:

[0006] Configuring a plurality of adjacent spiking rearrangement residual modules in a spiking rearrangement deep residual neural network;

[0007] Using the spiking rearrangement deep residual neural network after configuring the spiking rearrangement residual blocks as a first spiking rearrangement model;

[0008] Training the first spiking rearrangement model;

[0009] Inputting target data into the trained first spiking rearrangement model for processing to obtain a target processing result.

[0010] In some embodiments of the present invention, the first pulse rearrangement model is further provided with a first convolutional layer, a pooling layer, and a fully connected layer; processing the target data by inputting it into the trained first pulse rearrangement model includes:

[0011] Inputting the target data into the first convolutional layer for downsampling;

[0012] Inputting the downsampled target data into the plurality of adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result;

[0013] Sequentially inputting the first processing result into the pooling layer and the fully connected layer to obtain a second processing result.

[0014] In some embodiments of the present invention, obtaining the target processing result includes:

[0015] Obtaining the category corresponding to the target data according to the second processing result;

[0016] Obtaining a regression sequence and / or a regression single vector based on the second processing result;

[0017] Taking the category corresponding to the target data, the regression sequence and / or the regression single vector as the target processing result.

[0018] In some embodiments of the present invention, the pulse rearrangement residual module is sequentially configured with a pulse rearrangement layer, a second convolutional layer, a normalization layer, a pulse neuron layer, and a pulse anti-rearrangement layer; the step of inputting the downsampled target data into the plurality of adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result includes:

[0019] Inputting the downsampled target data into the pulse rearrangement layer to obtain a pulse rearrangement result;

[0020] Sequentially processing the pulse rearrangement result through the second convolutional layer, the normalization layer, the pulse neuron layer, and the pulse anti-rearrangement layer to obtain a first processing result.

[0021] In some embodiments of the present invention, the formula for inputting the downsampled target data into the pulse rearrangement layer to obtain a pulse rearrangement result is:

[0022]

[0023] where Y represents the rearrangement operation performed by the pulse rearrangement layer, X represents the downsampled target data; n, z, y, x represent the batch size, channel ordinal number, height, and width of the downsampled target data in sequence; M represents the total number of channels, r represents the rearrangement coefficient, % represents taking the remainder, represents rounding down.

[0024] In some embodiments of the present invention, the first pulse rearrangement residual module among the plurality of adjacent pulse rearrangement residual modules is further configured with a downsampling function.

[0025] In some embodiments of the present invention, before inputting the target data into the trained first pulse rearrangement model for processing to obtain the target processing result, it further includes:

[0026] Obtain the data to be input;

[0027] If the data to be input is a single digit, repeat the single digit a preset number of times to obtain a sequence with a preset length; if the data to be input is a sequence with a length of T composed of multiple digits, where T > 1, then no repetition is required;

[0028] Use the sequence with the preset length and the sequence with the length of T as the target data.

[0029] The second aspect of the present invention provides a data processing device based on a pulse rearrangement deep residual neural network, and the device includes:

[0030] A configuration module, configured to configure a plurality of adjacent pulse rearrangement residual modules in the pulse rearrangement deep residual neural network;

[0031] A rearrangement module, configured to use the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as the first pulse rearrangement model;

[0032] A training module, configured to train the first pulse rearrangement model;

[0033] A processing module, configured to input the target data into the trained first pulse rearrangement model for processing to obtain the target processing result.

[0034] The third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor runs the computer program to implement the following steps:

[0035] Configure a plurality of adjacent pulse rearrangement residual modules in the pulse rearrangement deep residual neural network;

[0036] Use the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as the first pulse rearrangement model;

[0037] Train the first pulse rearrangement model;

[0038] Input the target data into the trained first pulse rearrangement model for processing to obtain the target processing result.

[0039] A fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0040] Configure a plurality of adjacent pulse rearrangement residual modules in the pulse rearrangement deep residual neural network;

[0041] Use the pulse rearrangement deep residual neural network configured with the pulse rearrangement residual block as the first pulse rearrangement model;

[0042] Train the first pulse rearrangement model;

[0043] Input the target data into the trained first pulse rearrangement model for processing to obtain the target processing result.

[0044] The technical solution provided in the embodiments of the present application has at least the following technical effects or advantages:

[0045] In the present application, a plurality of adjacent pulse rearrangement residual modules are first configured in the pulse rearrangement deep residual neural network, and the pulse rearrangement deep residual neural network configured with the pulse rearrangement residual block is used as the first pulse rearrangement model. The first pulse rearrangement model is trained, and the target data is input into the trained first pulse rearrangement model for processing to obtain the target processing result. The rearrangement processing greatly reduces the number of parameters, reduces the risk of overfitting, and also reduces the storage and calculation overhead, thereby improving the efficiency of data processing. At the same time, the present application can obtain the category corresponding to the target data, and can obtain a regression sequence or a regression single vector, so as to be applicable to a variety of application scenarios.

[0046] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present application. And throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0048] Figure 1 A schematic diagram of the steps of a data processing method based on a pulse rearrangement deep residual neural network in an exemplary embodiment of the present application is shown;

[0049] Figure 2 A flowchart of a data processing method based on a pulse rearrangement deep residual neural network in an exemplary embodiment of the present application is shown;

[0050] Figure 3Shows a schematic diagram of the pulse matrix rearrangement and anti-rearrangement process in an exemplary embodiment of the present application;

[0051] Figure 4 Shows a schematic diagram of the structure of a data processing device based on a pulse rearrangement deep residual neural network in an exemplary embodiment of the present application;

[0052] Figure 5 Shows a schematic diagram of the structure of a computer device provided in an exemplary embodiment of the present application;

[0053] Figure 6 Shows a schematic diagram of a storage medium provided in an exemplary embodiment of the present application. Detailed implementation manners

[0054] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present application. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present application. It is obvious to those skilled in the art that the present application can be implemented without one or more of these details. In other examples, some technical features well-known to those skilled in the art are not described to avoid confusing the present application.

[0055] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should also be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or combinations thereof.

[0056] Now, exemplary embodiments according to the present application will be described in more detail with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many different forms and should not be construed as being limited only to the embodiments set forth herein. The drawings are not drawn to scale, and some details may be enlarged for the purpose of clear expression, and some details may be omitted. The shapes of various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are only exemplary. In practice, there may be deviations due to manufacturing tolerances or technical limitations, and those skilled in the art can design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0057] The following is combined with the specification appendix Figure 1 - Appendix Figure 6Several embodiments are given to describe the exemplary embodiments according to the present application. It should be noted that the following application scenarios are only shown for the convenience of understanding the spirit and principle of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0058] Embodiment 1:

[0059] This embodiment provides a data processing method based on a pulse rearrangement deep residual neural network, as Figure 1 shown, the method includes:

[0060] S1. Configure a plurality of adjacent pulse rearrangement residual modules in the pulse rearrangement deep residual neural network;

[0061] S2. Use the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as the first pulse rearrangement model;

[0062] S3. Train the first pulse rearrangement model;

[0063] S4. Input the target data into the trained first pulse rearrangement model for processing to obtain the target processing result.

[0064] In a specific implementation manner, referring to Figure 2 , the first pulse rearrangement model is also provided with a first convolutional layer, a pooling layer, and a fully connected layer; inputting the target data into the trained first pulse rearrangement model for processing includes: inputting the target data into the first convolutional layer for downsampling; inputting the downsampled target data into a plurality of adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result; inputting the first processing result into the pooling layer and the fully connected layer in sequence to obtain a second processing result.

[0065] In a specific implementation manner, obtaining the target processing result includes: obtaining the category corresponding to the target data according to the second processing result; obtaining a regression sequence and / or a regression single vector based on the second processing result; using the category corresponding to the target data, the regression sequence and / or the regression single vector as the target processing result.

[0066] In a specific implementation manner, the pulse rearrangement residual module is sequentially configured with a pulse rearrangement layer, a second convolutional layer, a normalization layer, a pulse neuron layer, and a pulse anti-rearrangement layer; inputting the downsampled target data into a plurality of adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result includes: inputting the downsampled target data into the pulse rearrangement layer to obtain a pulse rearrangement result; sequentially processing the pulse rearrangement result through the second convolutional layer, the normalization layer, the pulse neuron layer, and the pulse anti-rearrangement layer to obtain a first processing result.

[0067] In a specific implementation, the downsampled target data is input into a pulse rearrangement layer to obtain a pulse rearrangement result, which is represented by formula (1):

[0068]

[0069] Where Y represents the rearrangement operation performed by the pulse rearrangement layer, and X represents the downsampled target data; n, z, y, and x represent the batch size, channel ordinal number, height, and width of the downsampled target data in sequence; M represents the total number of channels, r represents the rearrangement coefficient, % represents taking the remainder, represents rounding down.

[0070] In a specific implementation, the first pulse rearrangement residual module among multiple adjacent pulse rearrangement residual modules is also configured with a downsampling function.

[0071] In a specific implementation, before the target data is input into the trained first pulse rearrangement model for processing to obtain the target processing result, it further includes: obtaining the data to be input; if the data to be input is a single digit, repeating the single digit a preset number of times to obtain a sequence with a preset length; if the data to be input is a sequence with a length of T composed of multiple digits, where T>1, no repetition is required; using the sequence with a preset length and the sequence with a length of T as the target data.

[0072] Embodiment 2:

[0073] This embodiment provides a data processing method based on a pulse rearrangement deep residual neural network. The steps included in the method are described in detail below.

[0074] The first step is to configure multiple adjacent pulse rearrangement residual modules in the pulse rearrangement deep residual neural network.

[0075] In a specific implementation, as Figure 2 shown, the pulse rearrangement deep residual neural network is provided with a first convolutional layer, a pooling layer, and a fully connected layer. The pooling layer pools to a 1*1 size and then performs a full connection through the fully connected layer. The pulse rearrangement residual module ( Figure 2 referred to as the pulse rearrangement element-wise residual block in [[ ]]) is sequentially configured with a pulse rearrangement layer, a second convolutional layer, a normalization layer, a pulse neuron layer, and a pulse anti-rearrangement layer. The pulse rearrangement layer, the second convolutional layer, the normalization layer, the pulse neuron layer, and the pulse anti-rearrangement layer can be regarded as stacked. There is no essential difference in the structures of the first convolutional layer and the second convolutional layer, except for distinguishing the convolutions at different positions. Multiple pulse rearrangement residual modules are arranged adjacent to each other as a stage, and the entire pulse rearrangement deep residual neural network can be configured with i (i>0) stages. The structure of multiple adjacent pulse rearrangement residual modules can be represented as:

[0076]

[0077] Among them, S[t] is the input of the entire pulse rearrangement residual module at time step t. is the i-th {pulse rearrangement-convolution-normalization-pulse neuron layer-pulse derangement} stack; n here is the total number of stacks (different from n in the fourth step). in formula (2) is the output obtained after the action of multiple {pulse rearrangement-convolution-normalization-pulse neuron layer-pulse derangement}, and together with the original input S[t], through the action of the connection function g, the output O[t] is obtained. The complete form of the connection function g is s o = g(s a , s b ), which is essentially an element-wise logical function. For the sake of distinction, the binary conversion of the values of s o to decimal is used to represent the corresponding g. For example, in the example shown in Table 1, s o takes values 1, 0, 1, 1 respectively, and the decimal conversion is 11, so the corresponding g is denoted as g 11 .

[0078] Table 1 Corresponding table of element-wise logical function values

[0079]

[0080]

[0081] According to the truth table composed of (s a , s b , s o ), 16 logical functions can be taken, and the corresponding g is {g0, g1,..., g 15}. It should be noted here that when training the network in the third step later, the gradient of g can be defined by numerical gradient. For example, using g 11 in Table 1, if it is required to solve , it can be solved using .

[0082] Step 2: Use the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual block as the first pulse rearrangement model.

[0083] Step 3: Train the first pulse rearrangement model.

[0084] During specific training, the parameters are the learning rate ∈, the network and its parameters θ, the training set D containing N data, the loss function and the total number of training epochs E.

[0085] [1] Let e = 1

[0086] [2] Let i = 1

[0087] [3] Take out (X, Y) = D[i] from the dataset, and obtain that the length of the input sequence is T

[0088] [4] Let t = 1

[0089] [5] Input X[t] into the network to get

[0090] [6] Let t = t + 1

[0091] [7] If t >, then go to [8]; otherwise go back to [5]

[0092] [8] Calculate the loss

[0093] [9] Backpropagate, perform gradient descent, and update the parameters

[0094]

[10] Let i = i + 1

[0095]

[11] If i >, then go to

[12] ; otherwise go back to [3]

[0096]

[12] Let e = e + 1

[0097]

[13] If e >, then exit; otherwise go back to [2]

[0098] If it is a data classification task, when the true class is j, for any t, Y[t][j] = 1; while for any k ≠ j, Y[t][k] = 0. The loss function can be the mean squared error or cross - entropy or other loss functions that measure distance. If it is a data regression task, when the regression target is a sequence, the loss function can directly be When the regression target is a single element Y, the loss function needs to be changed accordingly according to the type of output. When using the average value of all time steps as the regression result, the loss function can be When using the output at the last time step or the membrane potential of the last spiking neuron at the last time step as the regression result, the loss function can be

[0099] In the fourth step, input the target data into the trained first spiking rearrangement model for processing to obtain the target processing result

[0100] In one implementation, inputting target data into the trained first pulse rearrangement model includes: inputting the target data into a first convolutional layer for downsampling; inputting the downsampled target data into multiple adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result; and sequentially inputting the first processing result into a pooling layer and a fully connected layer to obtain a second processing result. The pulse rearrangement residual module is sequentially configured with a pulse rearrangement layer, a second convolutional layer, a normalization layer, a pulse neuron layer, and a pulse inverse rearrangement layer; inputting the downsampled target data into multiple adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result includes: inputting the downsampled target data into the pulse rearrangement layer to obtain a pulse rearrangement result; and sequentially processing the pulse rearrangement result through the second convolutional layer, the normalization layer, the pulse neuron layer, and the pulse inverse rearrangement layer to obtain a first processing result.

[0101] In a specific implementation, inputting the downsampled target data into the pulse rearrangement layer to obtain a pulse rearrangement result is represented by formula (1):

[0102]

[0103] where Y represents the rearrangement operation performed by the pulse rearrangement layer, X represents the downsampled target data; n, z, y, x represent the batch size, channel ordinal number, height, and width of the downsampled target data in sequence; M represents the total number of channels, r represents the rearrangement coefficient, % represents taking the remainder, represents rounding down.

[0104] Pulse rearrangement refers to splitting the data of the pulse on the channel into the width and height in sequence, while pulse inverse rearrangement is the reverse operation and is a special size transformation. For example, for an input pulse matrix with size [N, M, H, W] (where N is the batch size, M is the number of channels, H is the height, and W is the width) and a rearrangement coefficient of r 2 (where r is a positive integer), its shape is changed to Pulse inverse rearrangement is to perform the reverse operation and reshape the shape of [N, M, H, W] to Refer to Figure 3, the target data is four pulse matrices, namely a, b, c, and d, which are 4-channel pulse matrices composed of 2x2 pulse matrices, with a shape of [4, 2, 2]; after pulse rearrangement, a large 4x4 pulse matrix is obtained, with a size of [1, 4, 4]. The large pulse matrix of [1, 4, 4] is split into 4 channels to obtain a 4-channel pulse matrix of [4, 2, 2], which is the pulse inverse rearrangement. As a transformable implementation, if the target data is an image, after passing through the pulse rearrangement residual module, pixel rearrangement processing will be performed, followed by convolution processing, normalization processing, and pixel inverse rearrangement processing after passing through the pulse neuron layer, and finally pooling and fully connected layers. Additionally, if the input data is a single number, the single number is repeated a preset number of times to obtain a sequence with a preset length; if the input data is a sequence of multiple numbers with a length of T, where T > 1, then no repetition is required; the sequence with a preset length and the sequence with a length of T are used as the target data.

[0105] The function of the above normalization layer is to complete batch normalization or layer normalization. If batch normalization is used, the parameters of batch normalization can be merged with the convolutional layer to reduce the number of network parameters and increase the calculation speed. The specific method is as follows: Denote the weight of the convolution as W conv , and the bias is generally set to 0 because batch normalization has its own bias; Denote the weight of batch normalization as W bn , and the bias term is B bn , the mean value X of the statistical data m , the variance X v , then the weights and biases of the merged convolution are respectively:

[0106]

[0107] The pulse neuron layer refers to a layer composed of pulse neurons. The behavior of the pulse neuron layer can be described by three equations: charging, discharging, and resetting. Equation (3) represents the first equation, the charging equation:

[0108] H[t] = f(V[t - 1], X[t]) (3)

[0109] where X[t] is the input at time t. To avoid confusion, H[t] is used to represent the voltage after charging, and V[t] is used to represent the voltage after discharging. Here, f represents the charging equation, and different neurons have different charging equations. The charging equation is obtained by discretizing a continuous-time differential equation. For example, the subthreshold dynamics of an LIF neuron described by a continuous-time differential equation is represented by Equation (4):

[0110]

[0111] After discretization, the subthreshold discrete-time difference equation, which is the charging equation, is represented by Equation (5):

[0112]

[0113] where V rest is the resting potential and τ is the membrane time constant. The second equation is the firing equation, which is expressed by formula (6):

[0114] S[t] = Θ(H[t] - V th )(6)

[0115] S[t] is the pulse released by the neuron, Θ(x) is the Heaviside step function, which outputs 1 if and only if x ≥ 0, and outputs 0 otherwise. The firing equation indicates that when the voltage after the neuron is charged exceeds the threshold V th , a pulse 1 will be released, otherwise 0 is output. The third equation is the reset equation, which is expressed by formula (7):

[0116]

[0117] where, V reset represents the reset potential. Hard reset means hard reset, where the voltage is directly reset to V reset after the neuron releases a pulse. While Soft reset means soft reset, where after the neuron releases a pulse, the voltage will decrease by V th .

[0118] It should be noted that the derivative of Θ(x) in the firing equation is infinite when x = 0 and 0 when x ≠ 0. Using such a derivative directly for gradient descent will make the network unable to be trained. To solve this problem, the gradient substitution method is used. During forward propagation, Θ(x) is still used to ensure that the spiking neuron outputs a pulse; while during backpropagation, the derivative σ′(x) of the substitution function σ(x) is used. σ(x) is usually selected as a continuous function with a value range in (0,1), such as the common sigmoid function.

[0119] In a specific implementation manner, referring to Figure 2 again, obtaining the target processing result includes: obtaining the category corresponding to the target data according to the second processing result; obtaining a regression sequence and / or a regression single vector based on the second processing result; using the category corresponding to the target data, the regression sequence and / or the regression single vector as the target processing result. Among them, as Figure 2 shown, the output at the last moment of the model or the membrane potential at the last moment of the last layer of spiking neurons can be used as the regression single vector. It can be seen that the method described in this application is applicable to a variety of application scenarios.

[0120] Compared with ordinary convolution without using pulse rearrangement and anti-rearrangement, the number of parameters of this application is reduced from M in·M out ·K h ·K w Reduced to The number of parameters is greatly reduced, reducing the risk of overfitting and also reducing the storage and computational overhead.

[0121] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and do not limit the present invention.

[0122] Example 3:

[0123] This embodiment provides a data processing device based on a pulse rearrangement deep residual neural network, as Figure 4 shown, the device includes:

[0124] A configuration module 401, configured to configure a plurality of adjacent pulse rearrangement residual modules in a pulse rearrangement deep residual neural network;

[0125] A rearrangement module 402, configured to use the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as a first pulse rearrangement model;

[0126] A training module 403, configured to train the first pulse rearrangement model;

[0127] A processing module 404, configured to input target data into the trained first pulse rearrangement model for processing to obtain a target processing result.

[0128] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and do not limit the present invention.

[0129] It should also be emphasized that the system provided in the embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results of theory, methods, technologies, and application systems. Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technology mainly includes several major directions such as computer vision technology, robotics technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0130] Next, please refer to Figure 5 , which shows a schematic diagram of a computer device provided by some embodiments of the present application. As Figure 5As shown, the computer device 2 includes: a processor 200, a memory 201, a bus 202, and a communication interface 203. The processor 200, the communication interface 203, and the memory 201 are connected through the bus 202. A computer program that can run on the processor 200 is stored in the memory 201. When the processor 200 runs the computer program, it executes the data processing method based on the pulse rearrangement deep residual neural network provided in any of the foregoing embodiments of the present application.

[0131] Among them, the memory 201 may include a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 203 (which can be wired or wireless), a communication connection is realized between this system network element and at least one other network element, and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used.

[0132] The bus 202 can be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Among them, the memory 201 is used to store a program. After receiving an execution instruction, the processor 200 executes the program. The data processing method based on the pulse rearrangement deep residual neural network disclosed in any of the foregoing embodiments of the present application can be applied to the processor 200 or implemented by the processor 200.

[0133] The processor 200 may be an integrated circuit chip with the ability to process signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 200 or the instructions in the form of software. The above-mentioned processor 200 may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 201, and the processor 200 reads the information in the memory 201 and combines its hardware to complete the steps of the above method.

[0134] The embodiment of the present application also provides a computer-readable storage medium corresponding to the data processing method based on the pulse rearrangement deep residual neural network provided in the foregoing embodiment. Please refer to Figure 6 , Figure 6 The shown computer-readable storage medium is an optical disc 30, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will execute the data processing method based on the pulse rearrangement deep residual neural network provided in any of the foregoing embodiments.

[0135] In addition, examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here one by one.

[0136] The computer-readable storage medium provided in the above embodiments of the present application and the quantum key distribution channel allocation method in the space-division multiplexing optical network provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0137] The embodiments of the present application also provide a computer program product, including a computer program which, when executed by a processor, implements the steps of the data processing method based on the pulse rearrangement deep residual neural network provided by any of the foregoing embodiments, including: configuring a plurality of adjacent pulse rearrangement residual modules in the pulse rearrangement deep residual neural network; using the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as a first pulse rearrangement model; training the first pulse rearrangement model; and inputting target data into the trained first pulse rearrangement model for processing to obtain a target processing result.

[0138] It should be noted that: The algorithms and displays provided herein are not inherently related to any particular computer, virtual device, or other equipment. Various general-purpose devices can also be used in conjunction with the teachings provided herein. The structure required to construct such devices is obvious based on the above description. In addition, the present application is not directed to any specific programming language. It should be understood that the content of the present application described herein can be implemented using various programming languages, and the description of a specific language above is to disclose the best implementation mode of the present application. In the specification provided herein, a large number of specific details are set forth. However, it can be understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures, and technologies are not shown in detail so as not to obscure the understanding of this specification.

[0139] Similarly, it should be understood that, in order to streamline the present application and assist in understanding one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed present application requires more features than those expressly recited in each claim. Rather, as reflected in the following claims, the inventive aspects lie in less than all the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into the detailed description, where each claim itself serves as a separate embodiment of the present application.

[0140] Those skilled in the art can understand that the modules in the devices in the embodiments can be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition, they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all the features disclosed in this specification and all the processes or units of any method or device thus disclosed. Unless otherwise explicitly stated, each feature disclosed in this specification can be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0141] Each component embodiment of the present application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the virtual machine creation device according to the embodiments of the present application. The present application can also be implemented as a device or device program for executing part or all of the methods described herein. The program implementing the present application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0142] As described above, only the preferred specific embodiments of the present application are provided, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A data processing method based on a pulse rearrangement deep residual neural network, characterized in that The method includes: Configuring a plurality of adjacent pulse rearrangement residual modules in a pulse rearrangement deep residual neural network, where the pulse rearrangement residual modules are sequentially configured with a pulse rearrangement layer, a second convolutional layer, a normalization layer, a pulse neuron layer, and a pulse de-rearrangement layer; Regarding the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as a first pulse rearrangement model; Training the first pulse rearrangement model; Inputting target data into the trained first pulse rearrangement model for processing to obtain a target processing result, where the target data is an image; The first pulse rearrangement model is further provided with a first convolutional layer, a pooling layer, and a fully connected layer; the inputting the target data into the trained first pulse rearrangement model for processing includes: Inputting the target data into the first convolutional layer for downsampling; Inputting the downsampled target data into the plurality of adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result, including the formula: Among them, Y represents the rearrangement operation performed by the pulse rearrangement layer, and X represents the target data after downsampling; n, z, y, and x represent the batch size, channel ordinal number, height, and width of the target data after downsampling in sequence; M represents the total number of channels, r represents the rearrangement coefficient, % represents taking the remainder, denotes rounding down; Inputting the first processing result into the pooling layer and the fully connected layer in sequence to obtain a second processing result.

2. The data processing method based on the pulse rearrangement deep residual neural network according to claim 1, wherein The obtaining the target processing result includes: Obtaining the category corresponding to the target data according to the second processing result; Obtaining a regression sequence and / or a regression single vector based on the second processing result; Regarding the category corresponding to the target data, the regression sequence, and / or the regression single vector as the target processing result.

3. The data processing method based on the pulse rearrangement deep residual neural network according to claim 1, wherein The inputting the downsampled target data into the plurality of adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result includes: Inputting the downsampled target data into the pulse rearrangement layer to obtain a pulse rearrangement result; Subjecting the pulse rearrangement result to the processing of the second convolutional layer, the normalization layer, the pulse neuron layer, and the pulse de-rearrangement layer in sequence to obtain a first processing result.

4. The data processing method based on the pulse rearrangement deep residual neural network according to claim 1, wherein The first pulse rearrangement residual module among the plurality of adjacent pulse rearrangement residual modules is further configured with a downsampling function.

5. The data processing method based on the pulse rearrangement deep residual neural network according to claim 1, wherein Before the inputting the target data into the trained first pulse rearrangement model for processing to obtain a target processing result, it further includes: Obtaining the data to be input; If the data to be input is a single number, repeating the single number a preset number of times to obtain a sequence with a preset length; if the data to be input is a sequence with a length of T composed of multiple numbers, where T is greater than 1, then no repetition is required; Regarding the sequence with the preset length and the sequence with the length of T as the target data.

6. A data processing device based on a pulse rearrangement deep residual neural network, characterized in that, The device includes: A configuration module for configuring a plurality of adjacent pulse rearrangement residual modules in a pulse rearrangement deep residual neural network, where the pulse rearrangement residual modules are sequentially configured with a pulse rearrangement layer, a second convolutional layer, a normalization layer, a pulse neuron layer, and a pulse de-rearrangement layer; A rearrangement module for regarding the pulse rearrangement deep residual neural network after configuring the pulse rearrangement residual blocks as a first pulse rearrangement model; A training module for training the first pulse rearrangement model; A processing module for inputting target data into the trained first pulse rearrangement model for processing to obtain a target processing result, where the target data is an image; The first pulse rearrangement model is further provided with a first convolutional layer, a pooling layer, and a fully connected layer; the process of inputting the target data into the trained first pulse rearrangement model includes: Inputting the target data into the first convolutional layer for downsampling; Inputting the downsampled target data into the multiple adjacent pulse rearrangement residual modules for pulse rearrangement residual processing to obtain a first processing result, including the formula: Among them, Y represents the rearrangement operation performed by the pulse rearrangement layer, and X represents the target data after downsampling; n, z, y, and x respectively represent the batch size, channel ordinal number, height, and width of the target data after downsampling; M represents the total number of channels, r represents the rearrangement coefficient, % represents taking the remainder, represents rounding down; Sequentially inputting the first processing result into the pooling layer and the fully connected layer to obtain a second processing result.

7. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor runs the computer program to implement the steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-5.

Images