CN117875380A - Data processing method and device and electronic equipment - Google Patents
Data processing method and device and electronic equipment Download PDFInfo
- Publication number
- CN117875380A CN117875380A CN202211216171.9A CN202211216171A CN117875380A CN 117875380 A CN117875380 A CN 117875380A CN 202211216171 A CN202211216171 A CN 202211216171A CN 117875380 A CN117875380 A CN 117875380A
- Authority
- CN
- China
- Prior art keywords
- matrix
- network layer
- data
- target network
- data matrix
- 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.)
- Pending
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 24
- 239000011159 matrix material Substances 0.000 claims abstract description 562
- 238000012545 processing Methods 0.000 claims abstract description 104
- 238000000034 method Methods 0.000 claims abstract description 102
- 230000006870 function Effects 0.000 claims description 74
- 230000004913 activation Effects 0.000 claims description 23
- 238000007726 management method Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000004590 computer program Methods 0.000 description 5
- 238000002372 labelling Methods 0.000 description 4
- 230000003190 augmentative effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000013523 data management Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012958 reprocessing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
- 210000000707 wrist Anatomy 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application discloses a data processing method, a data processing device and electronic equipment, which belong to the technical field of data processing, and the data processing method in the embodiment of the application comprises the following steps: the method comprises the steps that electronic equipment obtains a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix; the electronic device performs a first process on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, and a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer.
Description
Technical Field
The application belongs to the technical field of data processing, and particularly relates to a data processing method, a data processing device and electronic equipment.
Background
In the conventional fully-connected network structure, if the input data is two-dimensional data, for example, the input data in the channel prediction field is two-dimensional data composed of a time dimension and an antenna dimension, or two-dimensional data composed of a time dimension and a frequency dimension, or the like, or the input data in the image processing field is two-dimensional image data, or the input data in the control field is two-dimensional control data (for example, data detected by a plurality of sensing devices, and data detected by each sensing device comprises sub-data of a plurality of dimensions), the two-dimensional data needs to be spliced into one vector for reprocessing, thus not only the distribution characteristic or the continuity of the data of at least one dimension can be hidden, but also the length of the input vector of the fully-connected network can be increased by many times, namely, the number of nodes of an input layer is very large, correspondingly, the length of the weight vector of the fully-connected network can be increased by many times, and the complexity of the structure of the fully-connected network and the computing complexity of the fully-connected network can be greatly increased due to the fact that the main operation of the fully-connected network depends on the weight, and the processing efficiency of the two-dimensional data is lower.
Disclosure of Invention
The embodiment of the application provides a data processing method, a data processing device and electronic equipment, which can solve the problem that the existing full-connection network has low efficiency of processing two-dimensional data.
In a first aspect, a data processing method is provided, the method comprising:
the method comprises the steps that electronic equipment obtains a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix;
the electronic device performs a first process on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1.
In a second aspect, there is provided a data processing apparatus comprising:
the acquisition module is used for acquiring a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix;
the processing module is used for executing first processing on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1.
In a third aspect, there is provided an electronic device comprising a processor and a memory storing a program or instructions executable on the processor, which when executed by the processor, implement the steps of the method as described in the first aspect.
In a fourth aspect, an electronic device is provided, including a processor and a communication interface, where the processor is configured to obtain a first data matrix, where the first data matrix includes a channel data matrix, an image data matrix, or a control data matrix; and executing first processing on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, and a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, wherein the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1.
In a fifth aspect, there is provided a readable storage medium having stored thereon a program or instructions which when executed by a processor realizes the steps of the method according to the first aspect.
In a sixth aspect, there is provided a chip comprising a processor and a communication interface coupled to the processor for running a program or instructions implementing the steps of the method according to the first aspect.
In a seventh aspect, there is provided a computer program/program product stored in a storage medium, the computer program/program product being executed by at least one processor to carry out the steps of the method according to the first aspect.
In the embodiment of the application, a first data matrix is acquired through electronic equipment, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix; the electronic device performs a first process on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, the target network layer is any network layer in the M network layers, namely, the fully-connected network provided by the application supports to directly process two-dimensional data such as a channel data matrix, an image data matrix or a control data matrix, so that distribution characteristics or continuity among various dimension data can be reserved.
Drawings
FIG. 1 is a flow chart of a data processing method according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a fully connected network according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of each network layer of a fully connected network according to an embodiment of the present application;
FIG. 4 is a block diagram of a data processing apparatus according to an embodiment of the present application;
fig. 5 is a block diagram of an electronic device provided in an embodiment of the present application;
fig. 6 is a block diagram of an electronic device according to an embodiment of the present application.
Detailed Description
Technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application are within the scope of the protection of the present application.
The terms first, second and the like in the description and in the claims, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application are capable of operation in sequences other than those illustrated or otherwise described herein, and that the terms "first" and "second" are generally intended to be used in a generic sense and not to limit the number of objects, for example, the first object may be one or more. Furthermore, in the description and claims, "and/or" means at least one of the connected objects, and the character "/" generally means a relationship in which the associated object is an "or" before and after.
The embodiment of the application provides a data processing method which can be applied to electronic equipment, wherein the electronic equipment can be a terminal, network side equipment or a server and the like.
The terminal may be a mobile phone, a tablet (Tablet Personal Computer), a Laptop (Laptop Computer) or a terminal-side Device called a notebook, a personal digital assistant (Personal Digital Assistant, PDA), a palm top, a netbook, an ultra-mobile personal Computer (ultra-mobile personal Computer, UMPC), a mobile internet appliance (Mobile Internet Device, MID), an augmented reality (augmented reality, AR)/Virtual Reality (VR) Device, a robot, a Wearable Device (weather Device), a vehicle-mounted Device (VUE), a pedestrian terminal (PUE), a smart home (home Device with a wireless communication function, such as a refrigerator, a television, a washing machine, or furniture), a game machine, a personal Computer (personal Computer, PC), a teller machine, or a self-service machine, and the Wearable Device includes: intelligent wrist-watch, intelligent bracelet, intelligent earphone, intelligent glasses, intelligent ornament (intelligent bracelet, intelligent ring, intelligent necklace, intelligent anklet, intelligent foot chain etc.), intelligent wrist strap, intelligent clothing etc.. It should be noted that, the embodiment of the present application is not limited to a specific type of terminal.
The network-side device may include an access network device or a core network device, where the access network device may also be referred to as a radio access network device, a radio access network (Radio Access Network, RAN), a radio access network function, or a radio access network element. The access network device may include a base station, a WLAN access point, a WiFi node, or the like, where the base station may be referred to as a node B, an evolved node B (eNB), an access point, a base transceiver station (Base Transceiver Station, BTS), a radio base station, a radio transceiver, a basic service set (Basic Service Set, BSS), an extended service set (Extended Service Set, ESS), a home node B, a home evolved node B, a transmission receiving point (Transmitting Receiving Point, TRP), or some other suitable terminology in the field, and the base station is not limited to a specific technical vocabulary so long as the same technical effect is achieved, and it should be noted that in the embodiment of the present application, only the base station in the NR system is described by way of example, and the specific type of the base station is not limited. The core network device may include, but is not limited to, at least one of: core network nodes, core network functions, mobility management entities (Mobility Management Entity, MME), access mobility management functions (Access and Mobility Management Function, AMF), session management functions (Session Management Function, SMF), user plane functions (User Plane Function, UPF), policy control functions (Policy Control Function, PCF), policy and charging rules function units (Policy and Charging Rules Function, PCRF), edge application service discovery functions (Edge Application Server Discovery Function, EASDF), unified data management (Unified Data Management, UDM), unified data repository (Unified Data Repository, UDR), home subscriber server (Home Subscriber Server, HSS), centralized network configuration (Centralized network configuration, CNC), network storage functions (Network Repository Function, NRF), network opening functions (Network Exposure Function, NEF), local NEF (or L-NEF), binding support functions (Binding Support Function, BSF), application functions (Application Function, AF), and the like.
The data processing method provided by the embodiment of the application is described in detail below by some embodiments and application scenarios thereof with reference to the accompanying drawings.
Referring to fig. 1, fig. 1 is a flowchart of a data processing method according to an embodiment of the present application, as shown in fig. 1, including the following steps:
step 101, an electronic device acquires a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix.
The number of rows and columns of the first data matrix may be greater than 1, that is, the first data matrix is two-dimensional data.
The channel data matrix may include, but is not limited to, a channel data matrix composed of a time dimension and an antenna dimension, or a channel data matrix composed of a time dimension or a frequency dimension, etc. For example, for a channel data matrix formed by a time dimension and an antenna dimension, if the number of antennas is N, the number of slots is L, where N and L are integers greater than 1, each slot corresponds to channel data from N antennas to a receiving end, that is, channel data corresponding to each slot is an N-dimensional vector h, and accordingly, the channel data matrix may be expressed as X:
wherein, h is as above n,i Representing channel data from the nth antenna to the receiving end during the ith time slot, wherein the value range of i is [1, L]The value range of n is [1, N]。
The image data matrix may include, but is not limited to, an image data matrix composed of image data of the same image, an image data matrix composed of image data of a plurality of images, or the like. The control data matrix may include, but is not limited to, a control data matrix composed of data of a plurality of sensors or data of an instrument or the like, or a control data matrix composed of data of a plurality of sensors or data of an instrument or the like at a plurality of times, wherein the data of the sensors or the data of the instrument or the like may include a plurality of sub-data, and for example, the data of each gyroscope may include angular velocity values of a plurality of directions.
It should be noted that, in the case where the first data matrix is a channel data matrix, the electronic device may be a terminal or a network side device; in the case where the first data matrix is an image data matrix or a control data matrix, the electronic device may be a terminal or a server.
Step 102, the electronic device performs a first process on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, where the fully-connected network includes M network layers, an input matrix of a first network layer of the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer of the M network layers, and a target network layer is configured to calculate based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer, and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, where the target network layer is any network layer of the M network layers, and M is an integer greater than or equal to 1.
It may be understood that, in this embodiment, the target network layers of the M network layers are taken as an example for explanation, and for each of the M network layers, the input matrix may be processed by using the same processing manner of the target network layer to obtain the output matrix, that is, the structure of each of the M network layers may be the same as that of the target network layer.
Specifically, when M is 1, the first network layer of the M network layers and the last network layer of the M network layers, and the target network layer are the same network layer, that is, the input matrix of the target network layer is the first data matrix, the target network layer may calculate based on the first data matrix, the weight matrix, and the offset matrix to obtain an output matrix, and the processing result is determined according to the output matrix of the target network layer.
When M is greater than 1, the M network layers are sequentially connected, for example, when the M network layers include a first network layer, a second network layer, and a third network layer, an output end of the first network layer is connected to an input end of the second network layer, an output end of the second network layer is connected to an input end of the third network layer, wherein the first network layer is a first network layer of the M network layers, and the third network layer is a last network layer of the M network layers. Optionally, in this case, each of the M network layers may calculate to obtain an output matrix of the network layer based on the input matrix of the network layer, the weight matrix corresponding to the network layer, and the offset matrix corresponding to the network layer, that is, each of the M network layers supports direct processing of the data matrix.
It should be noted that the dimensions (i.e., the number of rows and the number of columns) of the input matrix and the dimensions of the output matrix of each of the M network layers may be set reasonably according to practical application requirements. In addition, the dimension of the input matrix of the first network layer of the M network layers may be determined according to the dimension of the first data matrix, and the dimension of the output matrix of the last network layer of the M network layers may be determined according to a processing target, for example, if the processing target predicts N channel data corresponding to each time slot of the K time slots, the dimension of the output matrix of the last network layer of the M network layers may be n×k.
The weight matrix corresponding to each of the M network layers may include at least one weight matrix. In addition, the dimension of the weight matrix corresponding to each network layer may be determined according to the dimension of the input matrix and the dimension of the output matrix of the network layer, and the dimension of the offset matrix corresponding to each network layer may be determined according to the dimension of the output matrix of the network layer. Optionally, the number of rows and the number of columns of at least one weight matrix in the weight matrix corresponding to the network layer are both greater than 1, for example, the weight matrix corresponding to the network layer a includes a weight matrix a and a weight matrix B, and the number of rows and the number of columns of at least one weight matrix in the weight matrix a and the weight matrix B are both greater than 1. The dimension of the offset matrix corresponding to each of the M network layers may be determined according to the dimension of the output matrix of the network layer.
It will be appreciated that the respective weight values within the weight matrix and the respective offset values within the offset matrix for each of the M network layers are determined during the fully connected network training process. For example, a plurality of training data matrices may be obtained, and labeling data corresponding to each training data matrix may be obtained, where the labeling data is used to indicate a true value corresponding to each training data matrix, for example, if each training data matrix is made up of N channel data corresponding to each time slot in L time slots, each labeling data may be made up of N channel data corresponding to each time slot in K time slots in an actual situation, so that the fully connected network is trained based on the plurality of training data matrices, the labeling data corresponding to each training data matrix, and a set loss function (i.e., a cost function) to determine each weight value in the weight matrix corresponding to each network layer and each offset value in the offset matrix, where the loss function may be reasonably set according to an actual requirement, for example, the loss function may be a normalized minimum mean square error function, a minimum variance function, or a cross entropy function.
For example, for channel data of historical L slots (slots), future K slots of channel data are predicted by the fully connected network, a loss value between N channel data corresponding to each of the K slots predicted by the training data matrix (i.e., actual channel data) and N channel data corresponding to each of the K slots marked by the fully connected network may be calculated based on a normalized minimum mean square error function, and the fully connected network may be iteratively trained based on the loss value until the fully connected network converges. The normalized minimum mean square error function may be as follows:
wherein f represents a loss value,representing channel data of an nth antenna to a receiving end during a predicted kth time slot, y n,k Representing channel data from the nth antenna to the receiving end during the marked or actual kth time slot, the value of k being in the range of [1, k ]]Value range of nIs [1, N]。
The first process may include, but is not limited to, a prediction process, a classification process, a clustering process, a regression process, or the like, and optionally, the first process may correspond to the first data matrix, for example, if the first data matrix includes a channel data matrix, the first process may be a prediction process, that is, performing channel prediction; if the first data matrix includes an image data matrix, the first processing may include a classification processing, a clustering processing, a prediction processing, or the like; if the first data matrix includes a control data matrix, the first process may include a classification process, a clustering process, a regression process, or the like
The data processing method provided by the embodiment of the application obtains a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix; the method comprises the steps that a first processing is carried out on a first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, the target network layer is any network layer in the M network layers, namely, the fully-connected network provided by the method supports to directly process two-dimensional data such as a channel data matrix, an image data matrix or a control data matrix, so that distribution characteristics or continuity among various dimensional data can be reserved, and compared with the prior art, the complexity and the computational complexity of the fully-connected network can be reduced, and the efficiency of two-dimensional data processing can be improved.
Optionally, the target network layer is configured to: determining a second data matrix based on the input matrix of the target network layer and a weight matrix corresponding to the target network layer, and adding the second data matrix and an offset matrix corresponding to the target network layer to obtain a third data matrix;
the output matrix of the target network layer is the third data matrix or the data matrix of the third data matrix processed by the first activation function.
Illustratively, in the case that the weight matrix corresponding to the target network layer includes a weight matrix, the target network layer is configured to multiply the input matrix of the target network layer with the weight matrix to obtain the second data matrix; and under the condition that the weight matrix corresponding to the target network layer comprises a plurality of weight matrices, the target network layer is used for multiplying the input matrix of the target network layer by each weight matrix in the plurality of weight matrices respectively to obtain the second data matrix.
After the second data matrix is obtained, the second data matrix and the offset matrix corresponding to the target network layer may be added to obtain a third data matrix, where it is understood that the dimensions of the offset matrix corresponding to the target network layer are the same. The output matrix of the target network layer may be the third data matrix directly, that is, the target network layer directly outputs and uploads the third data matrix, or the output matrix of the target network layer may be a data matrix processed by the third data matrix through a first activation function, that is, the target network layer is further provided with the first activation function, and processes the third data matrix based on the first activation function, and outputs the data matrix obtained after the processing, where the first activation function may include, but is not limited to, a modified linear unit (Rectified Linear Unit, reLU) function, a softmax function, or the like.
In some alternative embodiments, each of the M network layers is provided with an activation function, and the activation functions of different network layers may be different, for example, the activation functions of all network layers except the last network layer of the M network layers are ReLU functions, and the activation function of the last network layer of the M network layers is a softmax function. In some alternative embodiments, each of the M network layers is not provided with an activation function, or each of the M network layers except the last network layer is provided with an activation function, and the last network layer of the M network layers is not provided with an activation function.
Optionally, the second data matrix is a data matrix obtained by multiplying an input matrix of the target network layer by a weight matrix corresponding to the target network layer.
In this embodiment, the weight matrix corresponding to the target network layer only includes one weight matrix, in which case, the target network layer may directly multiply the input matrix of the target network layer with the weight matrix to obtain the second data matrix. It should be understood that the multiplication of the input matrix and the weight matrix may be the left multiplication of the input matrix and the weight matrix, or the right multiplication of the input matrix and the weight matrix, which is not limited in this embodiment.
Each network layer in the fully-connected network provided by the embodiment can directly multiply the input matrix of the network layer with the corresponding weight matrix to obtain the second data matrix, so that the data distribution characteristics of each dimension of the input matrix can be quickly learned, the training speed of the fully-connected network can be further improved, and the processing efficiency of the fully-connected network obtained by training on the two-dimensional data is improved.
Optionally, the weight matrix corresponding to the target network layer includes a first weight matrix and a second weight matrix, and the second weight matrix is a data matrix obtained by respectively right multiplying the first weight matrix and left multiplying the second weight matrix by a data matrix input by the target network layer.
In this embodiment, the weight matrix corresponding to the target network layer includes two weight matrices, namely a first weight matrix and a second weight matrix, and the target network layer is configured to right multiply the input matrix of the target network layer by the first weight matrix and left multiply the input matrix of the target network layer by the second weight matrix, for example, the target network layer may first right multiply the input matrix of the target network layer by the first weight matrix and right multiply the data matrix obtained by right multiply by the second weight matrix to obtain the second data matrix; or the target network layer may first multiply the input matrix of the target network layer by a second weight matrix, and multiply the data matrix obtained by the left multiplication by the first weight matrix to obtain the second data matrix.
According to the full-connection network, the first weight matrix and the second weight matrix are correspondingly arranged on each network layer, the input matrix is multiplied by the first weight matrix and the second weight matrix to obtain the second data matrix, and compared with the situation that one weight matrix is correspondingly arranged on each network layer, the full-connection network is beneficial to more quickly learning the data distribution characteristics of each dimension of the input matrix, the training speed of the full-connection network can be further improved, and the processing efficiency of the full-connection network obtained through training on two-dimensional data is improved.
Optionally, the number of rows of the first weight matrix is the same as the number of columns of the input matrix of the target network layer, and the number of columns of the first weight matrix is the same as the number of columns of the output matrix of the target network layer;
the number of rows of the second weight matrix is the same as the number of rows of the output matrix of the target network layer, and the number of columns of the second weight matrix is the same as the number of rows of the input matrix of the target network layer.
Embodiments of the present application are illustrated below in conjunction with fig. 2 and 3:
exemplary, as shown in FIG. 2, the fully-connected network comprises three network layers, namely a first network layer, a second network layer and a third network layer, which are connected in turn, each of which may be structured as shown in FIG. 3, wherein W L Representing a second weight matrix (i.e., a left-hand weight matrix) corresponding to each network layer, W R The method comprises the steps of representing a first weight matrix (namely, right multiplying weight matrix) corresponding to each network layer, B represents an offset matrix corresponding to each network layer, and ReLU represents an activation function corresponding to each network layer, wherein the first network layer and the second network layer are provided with RELU, the third network layer is not provided with ReLU, input data of a fully connected network is X (namely, the first data matrix), and the dimension is NxL, wherein:
1) The input matrix of the first network layer is X, the first network layerThe output matrix is: x is X 1 ＝ReLU(W L1 XW R1 +B 1 )。
Wherein if the dimension of the output matrix of the first network layer is M 1L ×M 1R W is then L1 Is of dimension M 1L ×N，W R1 Is L x M 1R ，B 1 Is of dimension M 1L ×M 1R 。
2) The input matrix of the second network layer is X 1 The output matrix of the second network layer is: x is X 2 ＝ReLU(W L2 X 1 W R2 +B 2 )。
Wherein the dimension of the output matrix of the second network layer is M 2L ×M 2R ，W L2 Is of dimension M 2L ×M 1L ，W R1 Is of dimension M 1R ×M 2R ，B 2 Is of dimension M 2L ×M 2R 。
3) The input matrix of the third network layer is X 2 The output matrix of the third network layer is: y=w L3 X 2 W R3 +B 3 。
Wherein the output data of the third network layer is the output data of the whole fully-connected network, and the dimension of the output matrix of the third network layer is NxK, W L3 Is of dimension N x M 2L ，W R3 Is of dimension M 2R ×K，B 3 Is N x K.
As can be seen from the above, the number of rows of the left-hand weight matrix of each network layer of the fully connected network provided by the embodiment of the present application is the same as the number of rows of the output matrix row of the network layer, the number of columns of the left-hand weight matrix of each network layer is the same as the number of rows of the input matrix of the network layer, the number of columns of the right-hand weight matrix of each network layer is the same as the number of columns of the input matrix of the network layer, and the number of columns of the right-hand weight matrix of each network layer is the same as the number of columns of the output matrix of the network layer.
Optionally, in the case that the target network layer is a first network layer of the M network layers, the input matrix of the target network layer is the first data matrix, and the output matrix of the target network layer is an input matrix of a subsequent network layer of the target network layer;
in the case that the target network layer is the ith network layer of the M network layers, the input matrix of the target network layer is the output matrix of the ith-1 th network layer of the target network layers, and the output matrix of the target network layer is the input matrix of the (i+1) th network layer of the target network layers, wherein i is an integer greater than 1 and less than M;
And under the condition that the target network layer is the last network layer in the M network layers, the input matrix of the target network layer is the output matrix of the previous network side of the target network layer, and the processing result is determined according to the output matrix of the target network layer.
Optionally, the first processing is prediction processing, the processing result is a prediction result, and the prediction result is an output matrix of a last network layer of the M network layers;
or alternatively
The first process is a classification process, the processing result is a classification result, the classification result is an output matrix of a last network layer of the M network layers, and a first activation function of the last network layer is a softmax function.
In an embodiment, the first processing is a prediction processing, where the processing result is a prediction result, that is, the electronic device performs, through a fully-connected network obtained by training in advance, the prediction processing on the first data matrix to obtain a prediction result, where the prediction result may be output data of a last network layer (i.e., the output matrix) of the M network layers, that is, each data value in the output matrix of the last network layer is used to represent a predicted data value. For example, in the case where the first data matrix is a channel data matrix, each data value in the output matrix of the last network layer is used to characterize predicted channel data; in the case where the first data matrix is an image data matrix, each data value in the output matrix of the last network layer is used to characterize predicted image data; in case the first data matrix is a control data matrix, each data value in the output matrix of the last network layer is used to characterize the predicted control data.
In another embodiment, the first processing is a classification processing, where the processing result is a classification result, that is, the electronic device performs, through a fully-connected network obtained by training in advance, classification processing on the first data matrix to obtain a classification result, where the classification result may be output data of a last network layer (i.e., the output matrix) of the M network layers, where a first activation function of the last network layer is a softmax function, that is, output data of the last network layer of the M network layers is a data matrix processed by the softmax function. It should be noted that, for the classification process, the dimension of the output data of the last network layer is set according to the type number, for example, if the type number is S, that is, the output data of the last network layer is divided into S categories, the dimension may be 1×s or s×1.
Optionally, the first data matrix includes the channel data matrix, the first process is a prediction process, and the processing result is a prediction result;
the channel data matrix is used for representing N pieces of channel data corresponding to each time unit in L time units, the prediction result is used for representing N pieces of channel data corresponding to each time unit in K time units obtained through prediction, the K time units are located behind the L time units, N and L are integers greater than 1, and K is an integer greater than or equal to 1.
The time unit may include, but is not limited to, milliseconds, seconds, sub-slots, time slots, frames or subframes, etc., and is described below by taking the time unit as a slot (slot) as an example.
For a large-scale Multiple-Input Multiple-Output (MIMO) system, there are N antennas at a transmitting end (i.e., a terminal or a network side device), and a frame structure of data transmission of the MIMO system is performed on the basis of slots, and each slot has a duration of 1ms. MIMO predicts channel data for K slots in the future through a fully connected network using historical L slots of channel data (the channel data for each slot is an N-dimensional vector h, i.e., h includes N channel data). The data matrix X formed by the channel data of the L slots may be expressed as:
X＝[h 1 h 2 … h L ]
wherein h is i ＝[h 1,i h 2,i … h N,i ] T Channel vector from N antennas of massive MIMO (multiple input multiple output) to receiving end for ith slot, h n,i Representing channel data from the nth antenna to the receiving end during the ith slot, further, X may be expressed as:
from the above, the input data of the fully connected network is two-dimensional data.
The output data Y of the fully-connected network may represent predicted channel data of K slots, and the dimension of Y is nxk, and Y may be represented as:
Wherein y is n,k Represents the channel data from the nth antenna to the receiving end in the predicted kth slot period, and the value range of k is [1, K]The value range of n is [1, N]。
Optionally, the first data matrix includes the image data matrix, the first process is a classification process, and the processing result is a classification result;
the image data matrix is a matrix formed by image data of a target image, and the classification result is used for representing the type of the target image.
In this embodiment, the target image may be any image to be classified, the electronic device inputs an image data matrix of the target image into a fully-connected network obtained by training in advance to perform classification processing, so as to obtain a classification result, where the classification result is used to characterize a type of the target image, for example, if a preset type includes a first type and a second type, the classification result includes a probability value of the first type and a probability value of the second type, and the type of the target image is one type with a larger probability value of the first type and the second type.
It should be noted that, in the data processing method provided in the embodiment of the present application, the execution body may be a data processing apparatus, or a control module in the data processing apparatus for executing the data processing method. In the embodiment of the present application, a data processing device is described by taking an example that the data processing device executes a data processing method.
Referring to fig. 4, fig. 4 is a block diagram of a data processing apparatus according to an embodiment of the present application, and as shown in fig. 4, a data processing apparatus 400 includes:
an acquisition module 401, configured to acquire a first data matrix, where the first data matrix includes a channel data matrix, an image data matrix, or a control data matrix;
the processing module 402 is configured to perform a first process on the first data matrix through a fully-connected network obtained by pre-training to obtain a processing result, where the fully-connected network includes M network layers, an input matrix of a first network layer of the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer of the M network layers, and a target network layer is configured to calculate based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer, and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, where the target network layer is any network layer of the M network layers, and M is an integer greater than or equal to 1.
Optionally, the target network layer is configured to: determining a second data matrix based on the input matrix of the target network layer and a weight matrix corresponding to the target network layer, and adding the second data matrix and an offset matrix corresponding to the target network layer to obtain a third data matrix;
The output matrix of the target network layer is the third data matrix or the data matrix of the third data matrix processed by the first activation function.
Optionally, the second data matrix is a data matrix obtained by multiplying an input matrix of the target network layer by a weight matrix corresponding to the target network layer.
Optionally, the weight matrix corresponding to the target network layer includes a first weight matrix and a second weight matrix, and the second weight matrix is a data matrix obtained by respectively right multiplying the first weight matrix and left multiplying the second weight matrix by a data matrix input by the target network layer.
Optionally, the number of rows of the first weight matrix is the same as the number of columns of the input matrix of the target network layer, and the number of columns of the first weight matrix is the same as the number of columns of the output matrix of the target network layer;
the number of rows of the second weight matrix is the same as the number of rows of the output matrix of the target network layer, and the number of columns of the second weight matrix is the same as the number of rows of the input matrix of the target network layer.
Optionally, in the case that the target network layer is a first network layer of the M network layers, the input matrix of the target network layer is the first data matrix, and the output matrix of the target network layer is an input matrix of a subsequent network layer of the target network layer;
In the case that the target network layer is the ith network layer of the M network layers, the input matrix of the target network layer is the output matrix of the ith-1 th network layer of the target network layers, and the output matrix of the target network layer is the input matrix of the (i+1) th network layer of the target network layers, wherein i is an integer greater than 1 and less than M;
and under the condition that the target network layer is the last network layer in the M network layers, the input matrix of the target network layer is the output matrix of the previous network side of the target network layer, and the processing result is determined according to the output matrix of the target network layer.
Optionally, the first processing is prediction processing, the processing result is a prediction result, and the prediction result is an output matrix of a last network layer of the M network layers;
or alternatively
The first process is a classification process, the processing result is a classification result, the classification result is an output matrix of a last network layer of the M network layers, and a first activation function of the last network layer is a softmax function.
Optionally, the first data matrix includes the channel data matrix, the first process is a prediction process, and the processing result is a prediction result;
The channel data matrix is used for representing N pieces of channel data corresponding to each time unit in L time units, the prediction result is used for representing N pieces of channel data corresponding to each time unit in K time units obtained through prediction, the K time units are located behind the L time units, N and L are integers greater than 1, and K is an integer greater than or equal to 1.
Optionally, the first data matrix includes the image data matrix, the first process is a classification process, and the processing result is a classification result;
the image data matrix is a matrix formed by image data of a target image, and the classification result is used for representing the type of the target image.
The data processing apparatus in the embodiments of the present application may be an electronic device, for example, an electronic device with an operating system, or may be a component in an electronic device, for example, an integrated circuit or a chip. The electronic device may be a terminal, or may be other devices than a terminal. By way of example, the terminals may include, but are not limited to, the types of terminals listed above, and the other devices may be network side devices, servers, network attached storage (Network Attached Storage, NAS), etc., and embodiments of the present application are not limited in detail.
The data processing device provided in this embodiment of the present application can implement each process implemented by the method embodiment of fig. 1, and achieve the same technical effects, so that repetition is avoided, and no further description is given here.
Optionally, as shown in fig. 5, the embodiment of the present application further provides an electronic device 500, including a processor 501 and a memory 502, where the memory 502 stores a program or an instruction that can be executed on the processor 501, and the program or the instruction implements each step of the embodiment of the data processing method when executed by the processor 501, and the steps achieve the same technical effect, so that repetition is avoided, and no further description is given here.
The embodiment of the application also provides electronic equipment, which comprises a processor and a communication interface, wherein the processor is used for acquiring a first data matrix, and the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix; and executing first processing on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, and a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, wherein the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1. The implementation processes and implementation manners of the method embodiment are applicable to the electronic device embodiment, and the same technical effects can be achieved. Specifically, fig. 6 is a schematic hardware structure of an electronic device implementing an embodiment of the present application.
The electronic device 600 includes, but is not limited to: at least some of the components of the radio frequency unit 601, the network module 602, the audio output unit 603, the input unit 604, the sensor 605, the display unit 606, the user input unit 607, the interface unit 608, the memory 609, and the processor 610, etc.
Those skilled in the art will appreciate that the electronic device 600 may further include a power source (e.g., a battery) for powering the various components, which may be logically connected to the processor 610 by a power management system to perform functions such as managing charge, discharge, and power consumption by the power management system. The electronic device structure shown in fig. 6 does not constitute a limitation of the electronic device, and the electronic device may include more or less components than shown, or may combine certain components, or may be arranged in different components, which are not described in detail herein.
It should be understood that in the embodiment of the present application, the input unit 604 may include a graphics processing unit (Graphics Processing Unit, GPU) 6041 and a microphone 6042, and the graphics processor 6041 processes image data of still pictures or video obtained by an image capturing apparatus (such as a camera) in a video capturing mode or an image capturing mode. The display unit 606 may include a display panel 6061, and the display panel 6061 may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like. The user input unit 607 includes at least one of a touch panel 6071 and other input devices 6072. The touch panel 6071 is also called a touch screen. The touch panel 6071 may include two parts of a touch detection device and a touch controller. Other input devices 6072 may include, but are not limited to, a physical keyboard, function keys (e.g., volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, and so forth, which are not described in detail herein.
In this embodiment, after receiving downlink data from the network side device, the radio frequency unit 601 may transmit the downlink data to the processor 610 for processing; in addition, the radio frequency unit 601 may send uplink data to the network side device. Typically, the radio frequency unit 601 includes, but is not limited to, an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, and the like.
The memory 609 may be used to store software programs or instructions and various data. The memory 609 may mainly include a first storage area storing programs or instructions and a second storage area storing data, wherein the first storage area may store an operating system, application programs or instructions (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like. Further, the memory 609 may include volatile memory or nonvolatile memory, or the memory 609 may include both volatile and nonvolatile memory. The nonvolatile Memory may be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable EPROM (EEPROM), or a flash Memory. The volatile memory may be random access memory (Random Access Memory, RAM), static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (ddr SDRAM), enhanced SDRAM (Enhanced SDRAM), synchronous DRAM (SLDRAM), and Direct RAM (DRRAM). Memory 609 in the present embodiment includes, but is not limited to, these and any other suitable types of memory.
The processor 610 may include one or more processing units; optionally, the processor 610 integrates an application processor that primarily processes operations involving an operating system, user interface, application programs, etc., and a modem processor that primarily processes wireless communication signals, such as a baseband processor. It will be appreciated that the modem processor described above may not be integrated into the processor 610.
Wherein the processor 610 is configured to obtain a first data matrix, where the first data matrix includes a channel data matrix, an image data matrix, or a control data matrix; and executing first processing on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, and a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, wherein the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1.
Optionally, the target network layer is configured to: determining a second data matrix based on the input matrix of the target network layer and a weight matrix corresponding to the target network layer, and adding the second data matrix and an offset matrix corresponding to the target network layer to obtain a third data matrix;
the output matrix of the target network layer is the third data matrix or the data matrix of the third data matrix processed by the first activation function.
Optionally, the second data matrix is a data matrix obtained by multiplying an input matrix of the target network layer by a weight matrix corresponding to the target network layer.
Optionally, the weight matrix corresponding to the target network layer includes a first weight matrix and a second weight matrix, and the second weight matrix is a data matrix obtained by respectively right multiplying the first weight matrix and left multiplying the second weight matrix by a data matrix input by the target network layer.
Optionally, the number of rows of the first weight matrix is the same as the number of columns of the input matrix of the target network layer, and the number of columns of the first weight matrix is the same as the number of columns of the output matrix of the target network layer;
the number of rows of the second weight matrix is the same as the number of rows of the output matrix of the target network layer, and the number of columns of the second weight matrix is the same as the number of rows of the input matrix of the target network layer.
Optionally, in the case that the target network layer is a first network layer of the M network layers, the input matrix of the target network layer is the first data matrix, and the output matrix of the target network layer is an input matrix of a subsequent network layer of the target network layer;
in the case that the target network layer is the ith network layer of the M network layers, the input matrix of the target network layer is the output matrix of the ith-1 th network layer of the target network layers, and the output matrix of the target network layer is the input matrix of the (i+1) th network layer of the target network layers, wherein i is an integer greater than 1 and less than M;
and under the condition that the target network layer is the last network layer in the M network layers, the input matrix of the target network layer is the output matrix of the previous network side of the target network layer, and the processing result is determined according to the output matrix of the target network layer.
Optionally, the first processing is prediction processing, the processing result is a prediction result, and the prediction result is an output matrix of a last network layer of the M network layers;
or alternatively
The first process is a classification process, the processing result is a classification result, the classification result is an output matrix of a last network layer of the M network layers, and a first activation function of the last network layer is a softmax function.
Optionally, the first data matrix includes the channel data matrix, the first process is a prediction process, and the processing result is a prediction result;
the channel data matrix is used for representing N pieces of channel data corresponding to each time unit in L time units, the prediction result is used for representing N pieces of channel data corresponding to each time unit in K time units obtained through prediction, the K time units are located behind the L time units, N and L are integers greater than 1, and K is an integer greater than or equal to 1.
Optionally, the first data matrix includes the image data matrix, the first process is a classification process, and the processing result is a classification result;
the image data matrix is a matrix formed by image data of a target image, and the classification result is used for representing the type of the target image.
The embodiment of the present application further provides a readable storage medium, where a program or an instruction is stored on the readable storage medium, and the program or the instruction when executed by a processor implement each process of the foregoing data processing method embodiment, or implement each process of the foregoing data processing method embodiment, and achieve the same technical effect, so that repetition is avoided, and no further description is given here.
Wherein the processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes computer readable storage medium such as computer readable memory ROM, random access memory RAM, magnetic or optical disk, etc.
The embodiment of the present application further provides a chip, where the chip includes a processor and a communication interface, where the communication interface is coupled to the processor, and the processor is configured to run a program or an instruction, implement each process of the foregoing data processing method embodiment, or implement each process of the foregoing data processing method embodiment, and achieve the same technical effect, so that repetition is avoided, and no further description is provided herein.
It should be understood that the chips referred to in the embodiments of the present application may also be referred to as system-on-chip chips, or the like.
The embodiments of the present application further provide a computer program/program product, where the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement each process of the foregoing data processing method embodiment, or implement each process of the foregoing data processing method embodiment, and achieve the same technical effects, so that repetition is avoided, and no further description is given here.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. Furthermore, it should be noted that the scope of the methods and apparatus in the embodiments of the present application is not limited to performing the functions in the order shown or discussed, but may also include performing the functions in a substantially simultaneous manner or in an opposite order depending on the functions involved, e.g., the described methods may be performed in an order different from that described, and various steps may also be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solutions of the present application may be embodied essentially or in a part contributing to the prior art in the form of a computer software product stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk), comprising several instructions for causing a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method described in the embodiments of the present application.
The embodiments of the present application have been described above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those of ordinary skill in the art without departing from the spirit of the present application and the scope of the claims, which are also within the protection of the present application.
Claims (17)
1. A method of data processing, comprising:
the method comprises the steps that electronic equipment obtains a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix;
the electronic device performs a first process on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1.
2. The method of claim 1, wherein the target network layer is configured to: determining a second data matrix based on the input matrix of the target network layer and a weight matrix corresponding to the target network layer, and adding the second data matrix and an offset matrix corresponding to the target network layer to obtain a third data matrix;
the output matrix of the target network layer is the third data matrix or the data matrix of the third data matrix processed by the first activation function.
3. The method according to claim 2, wherein the second data matrix is a data matrix obtained by multiplying an input matrix of the target network layer by a weight matrix corresponding to the target network layer.
4. The method of claim 2, wherein the weight matrix corresponding to the target network layer includes a first weight matrix and a second weight matrix, and the second data matrix is a data matrix obtained by multiplying the first weight matrix and the second weight matrix by the data matrix input by the target network layer.
5. The method of claim 4, wherein the number of rows of the first weight matrix is the same as the number of columns of the input matrix of the target network layer, and the number of columns of the first weight matrix is the same as the number of columns of the output matrix of the target network layer;
The number of rows of the second weight matrix is the same as the number of rows of the output matrix of the target network layer, and the number of columns of the second weight matrix is the same as the number of rows of the input matrix of the target network layer.
6. The method according to claim 1, wherein, in case the target network layer is a first network layer of the M network layers, the input matrix of the target network layer is the first data matrix, and the output matrix of the target network layer is an input matrix of a subsequent network layer of the target network layer;
in the case that the target network layer is the ith network layer of the M network layers, the input matrix of the target network layer is the output matrix of the ith-1 th network layer of the target network layers, and the output matrix of the target network layer is the input matrix of the (i+1) th network layer of the target network layers, wherein i is an integer greater than 1 and less than M;
and under the condition that the target network layer is the last network layer in the M network layers, the input matrix of the target network layer is the output matrix of the previous network side of the target network layer, and the processing result is determined according to the output matrix of the target network layer.
7. The method according to any one of claims 1 to 5, wherein the first process is a prediction process, the process result is a prediction result, and the prediction result is an output matrix of a last network layer of the M network layers;
or alternatively
The first process is a classification process, the processing result is a classification result, the classification result is an output matrix of a last network layer of the M network layers, and a first activation function of the last network layer is a softmax function.
8. The method of claim 7, wherein the first data matrix comprises the channel data matrix, the first process is a predictive process, and the result of the process is a predictive result;
the channel data matrix is used for representing N pieces of channel data corresponding to each time unit in L time units, the prediction result is used for representing N pieces of channel data corresponding to each time unit in K time units obtained through prediction, the K time units are located behind the L time units, N and L are integers greater than 1, and K is an integer greater than or equal to 1.
9. The method of claim 7, wherein the first data matrix comprises the image data matrix, the first process is a classification process, and the processing result is a classification result;
The image data matrix is a matrix formed by image data of a target image, and the classification result is used for representing the type of the target image.
10. A data processing apparatus, comprising:
the acquisition module is used for acquiring a first data matrix, wherein the first data matrix comprises a channel data matrix, an image data matrix or a control data matrix;
the processing module is used for executing first processing on the first data matrix through a fully-connected network obtained through pre-training to obtain a processing result, wherein the fully-connected network comprises M network layers, an input matrix of a first network layer in the M network layers is the first data matrix, the processing result is determined according to an output matrix of a last network layer in the M network layers, a target network layer is used for calculating based on the input matrix of the target network layer, a weight matrix corresponding to the target network layer and an offset matrix corresponding to the target network layer to obtain an output matrix of the target network layer, the target network layer is any network layer in the M network layers, and M is an integer greater than or equal to 1.
11. The apparatus of claim 10, wherein the target network layer is configured to: determining a second data matrix based on the input matrix of the target network layer and a weight matrix corresponding to the target network layer, and adding the second data matrix and an offset matrix corresponding to the target network layer to obtain a third data matrix;
The output matrix of the target network layer is the third data matrix or the data matrix of the third data matrix processed by the first activation function.
12. The apparatus of claim 11, wherein the second data matrix is a data matrix obtained by multiplying an input matrix of the target network layer by a weight matrix corresponding to the target network layer.
13. The apparatus of claim 11, wherein the weight matrix corresponding to the target network layer comprises a first weight matrix and a second weight matrix, and the second data matrix is a data matrix obtained by multiplying the first weight matrix and the second weight matrix by the data matrix input by the target network layer.
14. The apparatus according to any one of claims 10 to 13, wherein the first process is a prediction process, the process result is a prediction result, and the prediction result is an output matrix of a last network layer of the M network layers;
or alternatively
The first process is a classification process, the processing result is a classification result, the classification result is an output matrix of a last network layer of the M network layers, and a first activation function of the last network layer is a softmax function.
15. The apparatus of claim 14, wherein the first data matrix comprises the channel data matrix, the first process is a predictive process, and the result of the process is a predictive result;
the channel data matrix is used for representing N pieces of channel data corresponding to each time unit in L time units, the prediction result is used for representing N pieces of channel data corresponding to each time unit in K time units obtained through prediction, the K time units are located behind the L time units, N and L are integers greater than 1, and K is an integer greater than or equal to 1.
16. An electronic device comprising a processor and a memory storing a program or instructions executable on the processor, which when executed by the processor, implement the steps of the data processing method of any one of claims 1 to 9.
17. A readable storage medium, characterized in that the readable storage medium has stored thereon a program or instructions which, when executed by a processor, implement the steps of the data processing method according to any of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211216171.9A CN117875380A (en) | 2022-09-30 | 2022-09-30 | Data processing method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211216171.9A CN117875380A (en) | 2022-09-30 | 2022-09-30 | Data processing method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117875380A true CN117875380A (en) | 2024-04-12 |
Family
ID=90579787
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211216171.9A Pending CN117875380A (en) | 2022-09-30 | 2022-09-30 | Data processing method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117875380A (en) |
-
2022
- 2022-09-30 CN CN202211216171.9A patent/CN117875380A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110162799B (en) | Model training method, machine translation method, and related devices and equipment | |
| CN109918684A (en) | Model training method, interpretation method, relevant apparatus, equipment and storage medium | |
| CN111816159B (en) | Language identification method and related device | |
| US11249645B2 (en) | Application management method, storage medium, and electronic apparatus | |
| CN113284142A (en) | Image detection method, image detection device, computer-readable storage medium and computer equipment | |
| CN112084959B (en) | Crowd image processing method and device | |
| CN110610191A (en) | Elevator floor identification method and device and terminal equipment | |
| CN116266216A (en) | Client screening method and device, client and central equipment | |
| CN112748899A (en) | Data processing method and related equipment | |
| CN117875380A (en) | Data processing method and device and electronic equipment | |
| CN116567806A (en) | Positioning method and communication equipment based on artificial intelligence AI model | |
| CN116415476A (en) | Model construction method and device and communication equipment | |
| CN112948763B (en) | Piece quantity prediction method and device, electronic equipment and storage medium | |
| CN110503189B (en) | Data processing method and device | |
| WO2023098535A1 (en) | Information interaction method and apparatus, and communication device | |
| WO2023098586A1 (en) | Information interaction method and apparatus, and communication device | |
| CN116450808B (en) | Data processing method and device and storage medium | |
| WO2024067665A1 (en) | Csi prediction processing method and apparatus, and communication device and readable storage medium | |
| WO2023098585A1 (en) | Information transmission method and apparatus, and communication device | |
| WO2024067280A1 (en) | Method and apparatus for updating ai model parameter, and communication device | |
| WO2024067281A1 (en) | Ai model processing method and apparatus, and communication device | |
| WO2024017176A1 (en) | Model training method and apparatus, network side device, and terminal device | |
| CN117093858A (en) | Model training processing method and device, terminal and network side equipment | |
| CN118042399A (en) | Model supervision method, terminal and network side equipment | |
| CN117896838A (en) | AI model reasoning method, apparatus and readable storage medium |
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 |