WO2023125880A1 - 模型的构建方法、装置及通信设备 - Google Patents

模型的构建方法、装置及通信设备 Download PDF

Info

Publication number
WO2023125880A1
WO2023125880A1 PCT/CN2022/143672 CN2022143672W WO2023125880A1 WO 2023125880 A1 WO2023125880 A1 WO 2023125880A1 CN 2022143672 W CN2022143672 W CN 2022143672W WO 2023125880 A1 WO2023125880 A1 WO 2023125880A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
communication device
configuration
model
prediction
Prior art date
Application number
PCT/CN2022/143672
Other languages
English (en)
French (fr)
Inventor
贾承璐
杨昂
孙鹏
Original Assignee
维沃移动通信有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 维沃移动通信有限公司 filed Critical 维沃移动通信有限公司
Publication of WO2023125880A1 publication Critical patent/WO2023125880A1/zh
Priority to US18/746,032 priority Critical patent/US20240346113A1/en

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/20Design optimisation, verification or simulation
    • G06F30/27Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217Validation; Performance evaluation; Active pattern learning techniques
    • G06F18/2193Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • G06N20/10Machine learning using kernel methods, e.g. support vector machines [SVM]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]

Definitions

  • the present application belongs to the technical field of wireless communication, and in particular relates to a method, device and communication equipment for constructing a model.
  • wireless communication networks may have various indicators such as high speed, low latency, and enhanced mobility.
  • NTN Non Terrestrial Networks
  • high-speed rail and other high-mobility scenarios due to the rapid changes in the surrounding scattering environment, the coherence time of the wireless channel is severely shortened, that is to say, wireless communication
  • the wireless channel detection results of equipment and network-side base stations may quickly fail, so more frequent wireless channel detection must be performed to maintain good communication performance, which poses a huge challenge to network load and energy consumption.
  • Time series prediction is a method for terminals and network-side devices to predict future communication measurements based on the observation of communication measurements in the past period of time. It is essentially based on historical measurement data to mine the time correlation between them.
  • time-series related information there are various types of terminals, different capabilities, and different requirements. Therefore, how to build a model for predicting time-series related information so that both parties in communication can understand and agree is a current need Solved technical problems.
  • the embodiments of the present application provide a model construction method, device, and communication device, which can solve the problem of how to construct a model for predicting time-series related information so that both communication parties can understand and agree.
  • a method for constructing a model including: a first communication device acquires configuration information of a time series prediction model from a second communication device, wherein the time series prediction model is used to predict time series related information; The first communication device constructs the time series prediction model by applying the configuration information.
  • an apparatus for constructing a model including: an acquisition module configured to acquire configuration information of a time-series prediction model from a second communication device, wherein the time-series prediction model is used to predict time-series-related information ; An application module, configured to apply the configuration information to construct the time series forecasting model.
  • a method for obtaining forecast information including: the second communication device configures configuration information of a time series forecast model for the first communication device, wherein the time series forecast model is used to predict time series related information
  • the second communication device receives the target prediction information reported by the first communication device, wherein the target prediction information is the prediction obtained by the first communication device using the time series prediction model to predict the first information information.
  • an apparatus for obtaining prediction information including: a configuration module, configured for a second communication device to configure configuration information of a time-series prediction model for the first communication device, wherein the time-series prediction model is used to predict time-series-related information; a receiving module, configured to receive target prediction information reported by the first communication device, wherein the target prediction information is the first information performed by the first communication device using the time-series prediction model Forecast information obtained by forecasting.
  • a communication device in a fifth aspect, includes a processor and a memory, the memory stores programs or instructions that can run on the processor, and when the programs or instructions are executed by the processor, the following is implemented: The steps of the method described in the first aspect, or the steps of implementing the method described in the third aspect.
  • a sixth aspect provides a communication device, including a processor and a communication interface, wherein the processor is configured to implement the steps of the method described in the first aspect, or implement the steps of the method described in the first aspect,
  • the communication interface is used for communicating with external equipment.
  • a model configuration system including: a first communication device and a second communication device, the first communication device can be used to perform the steps of the method described in the first aspect, and the second communication The device is operable to perform the steps of the method as described in the third aspect.
  • a readable storage medium is provided, and programs or instructions are stored on the readable storage medium, and when the programs or instructions are executed by a processor, the steps of the method described in the first aspect are realized, or the steps of the method described in the first aspect are realized, or The steps of the method described in the third aspect.
  • a ninth aspect provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, the processor is used to run programs or instructions, and implement the method as described in the first aspect steps, or to achieve the steps of the method as described in the third aspect.
  • a computer program/program product is provided, 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 the The steps of the method, or the steps of implementing the method as described in the third aspect.
  • the first communication device acquires the configuration information of the time series forecasting model used to predict time series related information from the second communication device, and applies the configuration information to construct the time series forecasting model, so that the first communication
  • the device can predict time-series-related information through the time-series prediction model, thereby solving the problem of how to configure a model for predicting time-series-related information so that both parties in communication can understand and agree.
  • FIG. 1 shows a block diagram of a wireless communication system to which an embodiment of the present application is applicable
  • Fig. 2 shows a schematic flow chart of a method for constructing a model provided by an embodiment of the present application
  • FIG. 3 shows a schematic flowchart of a method for obtaining prediction information provided by an embodiment of the present application
  • Fig. 4 shows another schematic flow chart of the method for constructing the model provided by the embodiment of the present application
  • FIG. 5 shows a schematic diagram of a training window and training sample sampling in an embodiment of the present application
  • Fig. 6 shows a schematic structural diagram of a model construction device provided by an embodiment of the present application
  • FIG. 7 shows a schematic structural diagram of a device for obtaining prediction information provided by an embodiment of the present application.
  • FIG. 8 shows a schematic structural diagram of a communication device provided by an embodiment of the present application.
  • FIG. 9 shows a schematic diagram of a hardware structure of a terminal provided by an embodiment of the present application.
  • FIG. 10 shows a schematic diagram of a hardware structure of a network side device provided by an embodiment of the present application.
  • first, second and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific sequence or sequence. 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 described herein and that "first" and “second” distinguish objects. It is usually one category, and the number of objects is not limited. For example, there may be one or more first objects.
  • “and/or” in the description and claims means at least one of the connected objects, and the character “/” generally means that the related objects are an "or” relationship.
  • LTE Long Term Evolution
  • LTE-Advanced LTE-Advanced
  • LTE-A Long Term Evolution-Advanced
  • CDMA Code Division Multiple Access
  • TDMA Time Division Multiple Access
  • FDMA Frequency Division Multiple Access
  • OFDMA Orthogonal Frequency Division Multiple Access
  • SC-FDMA Single-carrier Frequency Division Multiple Access
  • system and “network” in the embodiments of the present application are often used interchangeably, and the described technology can be used for the above-mentioned system and radio technology, and can also be used for other systems and radio technologies.
  • NR New Radio
  • the following description describes the New Radio (NR) system for example purposes, and uses NR terminology in most of the following descriptions, but these techniques can also be applied to applications other than NR system applications, such as the 6th generation (6 th Generation, 6G) communication system.
  • 6G 6th Generation
  • Fig. 1 shows a block diagram of a wireless communication system to which the embodiment of the present application is applicable.
  • the wireless communication system includes a terminal 11 and a network side device 12 .
  • the terminal 11 can be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer) or a notebook computer, a personal digital assistant (Personal Digital Assistant, PDA), a palmtop computer, a netbook, a super mobile personal computer (ultra-mobile personal computer, UMPC), mobile Internet device (Mobile Internet Device, MID), augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) equipment, robot, wearable device (Wearable Device) , vehicle equipment (VUE), pedestrian terminal (PUE), smart home (home equipment with wireless communication functions, such as refrigerators, TVs, washing machines or furniture, etc.), game consoles, personal computers (personal computers, PCs), teller machines or self-service Wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (
  • the network side device 12 may include an access network device and/or a core network device, wherein the access network device 12 may also be called a radio access network device, a radio access network (Radio Access Network, RAN), a radio access network function or radio access network unit.
  • RAN Radio Access Network
  • the access network device 12 may include a base station, a WLAN access point, or a WiFi node, etc., and the base station may be called a Node B, an evolved Node B (eNB), an access point, a Base Transceiver Station (Base Transceiver Station, BTS), a radio Base station, radio transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B, Home Evolved Node B, Transmission Reception Point (TRP) or all As long as the same technical effect is achieved, the base station is not limited to a specific technical vocabulary. It should be noted that in this embodiment of the application, only the base station in the NR system is used as an example for introduction, and The specific type of the base station is not limited.
  • FIG. 2 shows a schematic flow chart of a method for constructing a model provided by an embodiment of the present application. As shown in FIG. 2 , the method 200 mainly includes the following steps.
  • the first communication device acquires configuration information of a time series prediction model from the second communication device, where the time series prediction model is used to predict time series related information.
  • the first communication device and the second communication device may be two ends of wireless communication respectively, for example, the first communication device is a terminal, and the second communication device is a network side device.
  • the first communication device is a remote (Remote) terminal
  • the second communication device is a relay (Relay) terminal.
  • the first communication device and the second communication device may also be other communication devices, which are not specifically limited in this embodiment of the present application.
  • the network side equipment includes but is not limited to at least one of the following:
  • Core network nodes including NWDAF, LMF, etc., or neural network processing nodes.
  • Access network nodes including base stations or newly defined neural network processing nodes.
  • the first information predicted by the time series prediction model is time series related information
  • the first information may be channel state information (Channel State Information, CSI), location information of the first communication device
  • CSI Channel State Information
  • One or more items of information related to time series such as beam information of the first communication device.
  • the second communication device may actively send the above configuration information to the first communication device, for example, the second communication device actively sends the above configuration information to the first communication device according to the current application scenario of the first communication device.
  • Configuration information instructing the first communication device to deploy a time series prediction model to predict the first information.
  • the second communication device may also send the foregoing configuration information to the first communication device based on a request of the first communication device. Therefore, in this possible implementation manner, before the first communication device obtains the configuration information of the time series prediction model from the second communication device, the method further includes: the first communication device sends to the second communication device Model configuration request information. For example, the first communication device sends model configuration request information to the second communication device according to its capabilities, requesting the second communication device to configure the above configuration information for it.
  • the first communication device constructs the time series prediction model by applying the configuration information.
  • the first communication device constructs a time series prediction model according to the configuration information, so that the constructed time series prediction model can be used to predict the first information, so as to reduce the measurement of communication measurement quantities.
  • the first communication device acquires the configuration information of the time series forecasting model used to predict time series related information from the second communication device, and applies the configuration information to construct the time series forecasting model, so that the first communication
  • the device can predict time-series-related information through the time-series prediction model. Therefore, the problem of how to construct a model for predicting time series related information is solved so that both parties in communication can understand and agree.
  • the second communication device can The device can obtain the configuration information of the time series prediction model corresponding to the received prediction information, so that the second communication device can use the prediction information.
  • the first communication device is a terminal, since the capabilities of each terminal are different, by configuring the configuration information of the time series prediction model by the second communication device, an appropriate time series prediction model can be configured for each terminal.
  • the time series prediction model may be a Gaussian Process (Gaussian Process, GP) model.
  • the GP model is a machine learning method developed based on statistical learning theory and Bayesian theory. It can handle high-dimensional, small-sample and nonlinear complex regression and classification problems, and has strong generalization ability. Compared with methods such as neural networks, GP has the advantages of easy implementation, adaptive acquisition of hyperparameters, flexible non-parametric inference, interpretable and probabilistically meaningful output, etc., and has the potential to solve complex time series prediction problems in future wireless communication systems.
  • obeys 0 mean and the variance is Gaussian distribution:
  • the GP model estimates the parameter w by observing finite samples ⁇ x, y>, where x is the input of the GP model, and y is the label of the GP model.
  • the likelihood probability is expressed as follows:
  • the GP model is a supervised machine learning method that uses the maximum a posteriori probability criterion to infer model parameters based on observations of input X and output y:
  • the marginal probability has nothing to do with the parameter w and is a normalized constant:
  • the GP model takes into account the possibility of all w, and takes a weighted average of all possible linear model combinations.
  • the posterior distribution of the output also obeys the Gaussian distribution, and the mean value is used as the predicted value:
  • Predictive value It can be seen that the predicted value is the maximum likelihood estimate of the test input x * multiplied by the weight parameter w, namely The maximum likelihood estimate (noise-free) of .
  • the GP model can perform model training in the case of limited observation samples (x, y), and when the test data input x * is input to the GP model, it can give the label f * and prediction uncertainty (variance) at the same time. estimate.
  • the GP processing method first converts the nonlinear model in the low-dimensional space into a linear model in the high-dimensional space through the kernel function. , that is, through a set of basis functions ⁇ ( ), the sample vector x in a finite-dimensional space is mapped to a high-dimensional space, and the standard linear model is further extended as:
  • This model is a standard linear model after the expansion of the x dimension, and the processing method is the same as the standard linear model, that is, the predicted label f * in the case of the test input is x * obeys the following conditional distribution:
  • ⁇ (x) is a high-dimensional or even infinite-dimensional vector, and the computational complexity of the inner product of high-dimensional vectors is relatively high.
  • the kernel is introduced to approximate the inner product of high-dimensional vectors:
  • common kernels include: radial basis kernel:
  • ⁇ , l are hyperparameters. Due to the smoothness of infinite-order derivability, the radial basis kernel is the most widely used, and related theories prove that the radial basis kernel can be split into the inner product of two infinite-dimensional vectors, which can realize the transformation of the finite-dimensional vector x to the infinite-dimensional vector map.
  • test sample output follows a Gaussian distribution:
  • conditional Gaussian probability distribution can be obtained:
  • the mean value of the conditional Gaussian probability distribution is the maximum posterior probability estimate of the predicted value, and the variance represents the uncertainty of the prediction (variance):
  • n * and n are the sample numbers of the test set and the training set respectively.
  • the configuration information includes at least one of the following:
  • the first configuration information may include:
  • the number of kernel functions For example, the number of kernel functions used by the time series forecasting model.
  • the hyperparameters of the kernel function For example, the hyperparameters of the kernel function used by this time series forecasting model.
  • kernel function For example, which kernel functions are used in the time series forecasting model, such as square exponential kernel, quadratic rational kernel, constant kernel and other custom kernel functions.
  • a first task identifier wherein the first task identifier is used to indicate a task associated with the first configuration information.
  • the first task identifier indicates the task associated with the configuration of the kernel function in (1)
  • the time series prediction model is used to predict the first information
  • the first information may include CSI, position information, beam information, etc. related to time series information.
  • Optimizer selection information is used to optimize the hyperparameters in the kernel function.
  • optimizer selection information can include one of the following:
  • the first configuration parameters of the gradient descent optimization method wherein the first configuration parameters include: an initial value, a step size, and a termination condition. That is, the optimizer adopts the gradient descent optimization method, and the optimizer selection information includes the initial value setting, step size setting, and termination conditions in the gradient descent optimization method.
  • the second configuration parameters of the grid search optimization method wherein the second configuration parameters include: the upper bound of grid search with different hyperparameters, the lower bound of grid search with different hyperparameters, and the step of network search with different hyperparameters long. That is, the optimizer adopts the grid search method, and the optimizer selection information includes: the upper bound of grid search with different hyperparameters; the lower bound of grid search with different hyperparameters; the step size of grid search with different hyperparameters and other information.
  • the first model training configuration includes at least one of the following:
  • the length of the first training window is N time units.
  • Input for model training For example, the time unit number ⁇ 1, 2, ..., N ⁇ of the input for model training.
  • the time interval of the first training window sample sampling. That is, how often samples are collected in the first training window.
  • the first model predicts the configuration.
  • the first model prediction configuration may include at least one of the following:
  • the length of the first prediction window for example, if the CSI of M time units in the future is predicted based on the CSI of N time units, the length of the first prediction window is M time units.
  • the input of the prediction for example, based on the training data of N time units to predict the CSI of M time units in the future, the input of the model prediction is the time unit number corresponding to the CSI of M time units, for example, ⁇ N+1,... ,N+M ⁇ .
  • the predicted output of the time series prediction model may include: label information corresponding to the time unit number of the model prediction input, for example, ⁇ N+1,...,N+M ⁇ ; error information of the model prediction label, for example, The forecast variance output by the time series forecast model and/or the forecast error output by the time series forecast model.
  • the time interval of the first prediction window sample sampling. That is, the time between two samples in the first prediction window.
  • life cycle of the time series prediction model includes but is not limited to one of the following: effective time, failure time, and running time.
  • the computing mode configuration includes one of the following:
  • Serial computing that is, GP model training is performed on the same computing unit, for example, in a certain core or thread of a CPU.
  • Parallel computing that is, GP model training is performed on multiple computing units at the same time, such as in multiple cores or threads of a CPU.
  • the time unit in this embodiment of the present application is a preset unit time.
  • the time unit may include at least one of the following: reference signal period, prediction period, symbol (OFDM symbol), subframe, radio frame, Milliseconds, seconds, etc.
  • reference signal period prediction period
  • symbol OFDM symbol
  • subframe subframe
  • radio frame radio frame
  • Milliseconds milliseconds
  • the kernel function indicated by the first configuration information of the kernel function in (1) above is at least one kernel function in the kernel function list, wherein the first communication device and the second communication device
  • the kernel function list is pre-configured in the device.
  • the list of kernel functions may include identifiers of one or more kernel functions, and a kernel function may be uniquely determined through the identifiers.
  • the task identification in (2) above is at least one task in a task list, wherein the task list is pre-configured in the first communication device and the second communication device . That is to say, the first communication device configures the same task list related to time series prediction as the second communication device in advance, and the tasks can be uniquely determined by corresponding task identifiers.
  • the optimizer selection information in (3) above includes a target optimizer identifier, and the target optimizer identifier is at least one optimizer identifier in the optimizer list, and the optimizer list includes a correspondence between an optimizer identifier and a parameter configuration, and the optimizer list is pre-configured in the first communication device and the second communication device. That is to say, the first communication device configures the same optimizer selection and specific parameter configuration list as the second communication device in advance, and the optimizer selection and configuration information can be uniquely determined by the list identifier.
  • model training configuration in (4) above may also include:
  • a first processing method for training input wherein the first processing method includes: linear scaling processing or nonlinear scaling processing;
  • a second processing manner for labels includes: linear normalization processing (for example, scale*X) or mean value normalization processing (for example, log2(X)).
  • the computing mode configuration may be determined by the first communication device according to its computing capability and storage capability, or may be dynamically configured by the second communication device to the first communication device.
  • the method may further include: the first communication device sends feedback information to the second communication device , wherein the feedback information is used to indicate that the first communication device supports the configuration information, or the feedback information is used to indicate that the first communication device does not support the configuration information.
  • the first communication device may feed back to the second communication device whether the first communication device supports the configuration information, so that the second The communication device may acquire whether the first communication device can configure a time series prediction model corresponding to the configuration information.
  • the feedback information includes but is not limited to at least one of the following:
  • the second indication information is used to indicate kernel functions supported by the first communication device, where the kernel functions are one or more kernel functions configured in the configuration information; That is to say, the second indication information is used to indicate which one or more of the kernel functions configured by the configuration information are supported by the first communication device.
  • the third indication information wherein the third indication information is used to indicate that the first communication device has a task prediction requirement, or the third indication information is used to indicate that the first communication device does not have The predicted demand of the task, the task is a task configured in the configuration information; that is, the third indication information is used to indicate whether the first communication device has a predicted demand of a related task configured in the configuration information.
  • the fifth indication information is used to indicate that the first communication device supports the model training configuration in the configuration information, or the fifth indication information is used to indicate that the first communication device supports the model training configuration in the configuration information.
  • a communication device does not support the model training configuration in the configuration information; that is, the fifth indication information indicates whether the first communication device supports the model training configuration configured in the configuration information.
  • the first communication device may not support a longer training window size considering its limited storage capacity and/or computing capacity, if the length of the first training window configured by the configuration information exceeds the storage capacity and/or the first communication device's storage capacity and/or or computing capability, the first communication device may feed back that it does not support the model training configuration in the configuration information.
  • Sixth indication information wherein the sixth indication information is used to indicate that the first communication device supports the model prediction configuration in the configuration information, or the sixth indication information is used to indicate that the first communication device supports the model prediction configuration in the configuration information.
  • a communication device does not support the model prediction configuration in the configuration information. That is to say, the sixth indication information is used to indicate whether the first communication device supports the model prediction configuration in the configuration information, for example, whether it supports the length of the prediction window configured in the configuration information.
  • the The method may further include: the first communication device recommending the recommended configuration of the time series prediction model to the second communication device according to the second information.
  • the configuration information configured by the second communication device for the first communication device can be more adapted to the first communication device.
  • the recommended configuration includes at least one of the following:
  • a recommended kernel function that is, a kernel function recommended by the first communication device.
  • the second information includes at least one of the following:
  • Statistical information of the first information to be predicted for example, the mean value and/or variance of the first information
  • the estimation error of the first information refers to the estimation error of the known information (historical information).
  • the CSI of the past N time units is estimated by the terminal based on the pilot frequency (SRS), and there is an error in the estimation;
  • model prediction error for example, the mean and/or variance of the model prediction error.
  • Mobility information of the first communication device for example, at least one of the moving speed of the first communication device, beam switching information, cell switching information, and the like.
  • Statistical information of noise for example, signal-to-noise ratio, signal-to-interference-to-noise ratio, and the like.
  • Performance requirement information wherein the performance requirement information includes at least one of the following: prediction accuracy requirement information, processing delay requirement information, and calculation delay requirement information.
  • the first communication device after the first communication device applies the configuration information to configure the time series forecasting model, it may train the time series forecasting model based on the above model training configuration, and after the training, based on the above
  • the model forecasting configuration uses the time series forecasting model to forecast the first information.
  • the processing method of prediction input and output is the same as that of training input and output during training.
  • the method may further include: the first communication device reports the target prediction to the second communication device information, wherein the target prediction information is prediction information obtained by using the time series prediction model to predict the first information. That is, the first communication device reports the predicted first information to the second communication device. For example, according to the measured CSI information of N time units, the first communication device predicts the CSI information of M time units in the future through a time series prediction model, and reports the predicted CSI information of M time units.
  • the target prediction information includes at least one of the following:
  • a second task identifier is used to indicate a specific category of the first information, for example, CSI or beam information.
  • the second task identifier may be the above-mentioned first task identifier, or may be one of the above-mentioned first task identifiers.
  • the first communication device may have multiple prediction tasks at the same time, the second communication device configures a configuration information based on each prediction task, and the first communication device carries the second task identifier in the reported target prediction information to indicate the currently reported target prediction The prediction task corresponding to the information.
  • the second training window is the training window corresponding to the configuration of the training window configured in the configuration information It may be the same, or it may be one of the training windows corresponding to the training window configuration configured in the above configuration information.
  • the configuration-related information of the second training window may include at least one of the following:
  • the training data contains CSI of N time units, and N is the length of the training window.
  • first time stamp information includes at least one of the following: the number of the second training window, the start time of the second training window (that is, the start time), and the end time of the second training window (that is, the end time), and the time interval of sampling samples in the second training window (that is, the time interval between two samples in the second training window). For example, if the CSI of the second training window is the CSI of the 0th, 2nd, 4th, 6th, and 8th time units, then the time interval of sample sampling in the second training window is 2 time units.
  • the configuration-related information of the second prediction window includes at least one of the following:
  • second time stamp information includes at least one of the following: the number of the second prediction window, the start time of the second prediction window, the end time of the second prediction window, the second prediction window The interval at which samples are sampled.
  • the time series prediction model may be based on a neural network time series prediction model.
  • the configuration information includes at least one of the following:
  • the structural information of the time series forecasting model can include:
  • neural network for example, convolutional neural network, fully connected neural network, Transformer, etc.
  • hyperparameter information of the neural network, wherein the hyperparameter information may include:
  • the number of key modules for example, if the Transformer model is adopted, it may include the number of encoding (encoder) modules and/or the number of decoding (decoder) modules of Transformer.
  • the weight of the time series prediction model for example, the weight, bias, etc. of each node in the neural network
  • the configuration of the time series prediction model includes at least one of the following: an optimizer and a loss function; for example, by compiling some information of the model, such as an optimizer and a loss function.
  • the first communication device acquires configuration information of the time series prediction model from the second communication device, including one of the following:
  • the first communication device agrees with the second communication device on the configuration information, that is, statically configures the configuration information.
  • the first communication device determines the configuration information, and reports the configuration information to the second communication device, and acquires the configuration information after receiving confirmation information from the second communication device.
  • Fig. 3 shows a schematic flow chart of a method for obtaining prediction information provided by an embodiment of the present application. As shown in Fig. 3, the method 300 mainly includes:
  • the second communication device configures configuration information of a time series prediction model for the first communication device, where the time series prediction model is used to predict first information, and the first information is information related to time series.
  • the configuration information is the same as the configuration information in the method 200, and refer to the description in the method 200 for details.
  • the second communication device receives target prediction information reported by the first communication device, where the target prediction information is the first communication device predicting the first information using the time series prediction model The forecast information obtained.
  • Method 300 is an execution method corresponding to method 200 on the second communication device side, and has a possible implementation manner that is the same as or corresponding to method 200 . For details, refer to the description in method 200 . Only some possible implementations involved in the method 300 will be described below.
  • the configuration information includes at least one of the following:
  • the first configuration information of the kernel function includes: the number of kernel functions, the hyperparameters of the kernel function, and the type of the kernel function;
  • Model training configuration includes at least one of the following: the length of the first training window, the input of the model training, the label of the model training, the time interval of the first training window sample sampling;
  • Model prediction configuration wherein, the model prediction configuration includes at least one of the following: the length of the prediction window, the input of the prediction, the output of the prediction, the time interval of sampling of the prediction window samples;
  • the optimizer selection information includes one of the following:
  • the first configuration parameter of the gradient descent optimization method wherein the first configuration parameter includes: initial value, step size, and termination condition;
  • the second configuration parameter of the grid search optimization method wherein the second configuration parameter is: the upper bound of grid search with different hyperparameters, the lower bound of grid search with different hyperparameters, and the network search with different hyperparameters step size.
  • the predicted output includes:
  • the life cycle of the time series prediction model includes: effective time, invalid time, and running time.
  • the method further includes:
  • the second communication device receives feedback information sent by the first communication device, where the feedback information is used to indicate that the first communication device supports the configuration information, or the feedback information is used to indicate that the first communication device supports the configuration information.
  • a communication device does not support the configuration information.
  • the feedback information includes at least one of the following:
  • second indication information where the second indication information is used to indicate kernel functions supported by the first communication device, where the kernel functions are one or more kernel functions configured in the configuration information;
  • the third indication information wherein the third indication information is used to indicate that the first communication device has a task prediction requirement, or the third indication information is used to indicate that the first communication device does not have The demand predicted by the task, the task is the task configured by the configuration information;
  • the fifth indication information is used to indicate that the first communication device supports the model training configuration in the configuration information, or the fifth indication information is used to indicate that the first communication device supports the model training configuration in the configuration information.
  • a communication device does not support the model training configuration in the configuration information;
  • Sixth indication information wherein the sixth indication information is used to indicate that the first communication device supports the model prediction configuration in the configuration information, or the sixth indication information is used to indicate that the first communication device supports the model prediction configuration in the configuration information.
  • a communication device does not support the model prediction configuration in the configuration information.
  • the kernel function is at least one kernel function in a kernel function list, where the kernel function list is pre-configured in the first communication device and the second communication device.
  • the task identifier is at least one task in a task list, where the task list is pre-configured in the first communication device and the second communication device.
  • the optimizer selection information includes a target optimizer ID, and the target optimizer ID is at least one optimizer ID in an optimizer list, and the optimizer list includes the optimizer ID and Corresponding relationship of parameter configuration, the optimizer list is pre-configured in the first communication device and the second communication device.
  • model training configuration further includes:
  • a first processing method for training input wherein the first processing method includes: linear scaling processing or nonlinear scaling processing;
  • a second processing method for labels wherein the second processing method includes: linear normalization processing or mean value normalization processing.
  • the computing mode configuration is determined by the first communication device according to its computing capability and storage capability.
  • the method before the second communication device configures the configuration information of the time series prediction model for the first communication device, the method further includes: the second communication device receiving the information sent by the first communication device Recommended configurations for the time series forecasting models described above.
  • the recommended configuration includes at least one of the following:
  • the second communication device configures the configuration information of the time series prediction model for the first communication device, including:
  • the second communication device receives the second information reported by the first communication device
  • the second information includes at least one of the following:
  • Performance requirement information wherein the performance requirement information includes at least one of the following: prediction accuracy requirement information, processing delay requirement information, and calculation delay requirement information.
  • the target prediction information includes at least one of the following:
  • the relevant information of the second training window configuration includes at least one of the following:
  • First timestamp information wherein the first timestamp information includes at least one of the following: the number of the second training window, the start time of the second training window, the end time of the second training window, the second training window sample sampling time interval.
  • the relevant information of the second prediction window includes at least one of the following:
  • the second time stamp information includes at least one of the following: the number of the second prediction window, the start time of the second prediction window, the end time of the second prediction window, the second prediction window sample sampling time interval.
  • the method before the second communication device configures the configuration information of the time series prediction model for the first communication device, the method further includes: receiving model configuration request information sent by the first communication device.
  • the time series forecasting model includes: a time series forecasting model based on a neural network; wherein the configuration information includes at least one of the following:
  • the configuration of the time series forecasting model includes at least one of the following: optimizer, loss function;
  • the second communication device configures the configuration information of the time series prediction model for the first communication device, including one of the following:
  • the second communication device aperiodically configures the configuration information of the time series prediction model for the first communication device
  • the second communication device periodically configures the configuration information of the time series prediction model for the first communication device
  • the second communication device reports the configuration information, and sends confirmation information to the first communication device.
  • FIG. 4 shows another flow chart of the method for constructing a model provided by the embodiment of the present application. As shown in FIG. 3 , the method 400 mainly includes the following steps.
  • the terminal requests GP model configuration information from the core network device.
  • the requested GP model configuration information includes:
  • Kernel function configuration Radial basis kernel function and rational quadratic kernel function are used as the kernel function of CSI prediction:
  • Task ID: "1" is used to indicate CSI prediction.
  • Model training configuration the length of the training window is N time units, that is, the number of historical CSIs in the training window is N, the model training input is ⁇ 1, 2,...,N ⁇ , and the label of the model training is the time unit number ⁇ 1 ,2,...,N ⁇ corresponding to the CSI, the sampling time interval of training window samples is 2 time units;
  • Model prediction configuration information the length of the prediction window is M time units, that is, based on the CSI of N time units in the past, the CSI of M time units in the future is predicted, and the model test input is the time unit number ⁇ N+1, N+2, ...,N+M ⁇ , the prediction output is the CSI corresponding to the time unit number ⁇ N+1,N+2,...,N+M ⁇ and the corresponding prediction variance.
  • the length of the training window is 6 time slots
  • the sample sampling interval of the training window is 2 time slots
  • the length of the prediction window is 1 time slot
  • the prediction window The sampling interval is 2 time slots.
  • the GP model takes effect from the sth time unit, and expires at the s'th time unit.
  • the network side device configures GP model related information to the terminal.
  • the terminal reports GP model prediction information to the network side device.
  • the reported forecast information may include:
  • the reported training window configuration related information may include:
  • the length of the training window is 6 time units
  • the model training input is the time unit number ⁇ 1, 2, ..., 6 ⁇ ;
  • the timestamp information may include:
  • the number of the training window is T, which represents the relative time compared to the start of the training window
  • the start time of the training window is s, indicating the start time of the training
  • the end time of the training window is s’, indicating the end time of the training
  • the sampling time interval of training window samples is 2, and the CSI of ⁇ 0, 2, ..., 10, 12 ⁇ time units is used for training.
  • the prediction window configuration information may include:
  • the prediction window length is 1, based on the CSI of 6 time units of the training window, the CSI of 1 time unit in the future is predicted;
  • Timestamp information may include:
  • the number of the prediction window is T’, which indicates the relative time compared to the start of the prediction window
  • the start time of the prediction window is p, which represents the start time of the prediction
  • the end time of the prediction window is p’, which indicates the end time of the prediction
  • the sampling time interval of the prediction window sample is 2, and the CSI of the ⁇ 14 ⁇ th time unit is predicted.
  • the interactive configuration information and process are provided, so that the terminal can be deployed based on the configuration of the network side equipment Time series forecasting model, which predicts parameters related to time series.
  • the execution subject may be a device for constructing a model.
  • the model construction device provided in the embodiment of the present application is described by taking the model construction device executing the model construction method as an example.
  • FIG. 6 shows a schematic structural diagram of the model construction device provided by the embodiment of the present application.
  • the device 600 mainly includes: an acquisition module 601 and an application module 602 .
  • the obtaining module 601 is configured to obtain configuration information of a time series prediction model from the second communication device, wherein the time series prediction model is used to predict time series related information; the application module 602 is used to Applying the configuration information to construct the time series forecasting model.
  • the apparatus further includes: a first sending module, configured to send feedback information to the second communication device, where the feedback information is used to indicate that the first communication device supports the The configuration information or the feedback information is used to indicate that the first communication device does not support the configuration information.
  • a first sending module configured to send feedback information to the second communication device, where the feedback information is used to indicate that the first communication device supports the The configuration information or the feedback information is used to indicate that the first communication device does not support the configuration information.
  • the apparatus further includes: a recommendation module, configured to recommend a recommended configuration of the time series prediction model to the second communication device according to the second information.
  • the apparatus further includes: a reporting module, configured to report target prediction information to the second communication device, where the target prediction information is the The predicted information obtained by predicting the information.
  • the device for constructing the model in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or a component in the electronic device, such as an integrated circuit or a chip.
  • the electronic device may be a terminal, or other devices other than the terminal.
  • the terminal may include, but not limited to, the types of terminal 11 listed above, and other devices may be servers, Network Attached Storage (NAS), etc., which are not specifically limited in this embodiment of the present application.
  • NAS Network Attached Storage
  • the device for constructing the model provided by the embodiment of the present application can implement various processes implemented by the first communication device or terminal in the method embodiments in FIGS. 2 to 5 , and achieve the same technical effect. To avoid repetition, details are not repeated here.
  • FIG. 7 shows a schematic structural diagram of an apparatus for obtaining prediction information provided by an embodiment of the present application.
  • the apparatus 700 mainly includes: a configuration module 701 and a receiving module 702 .
  • the configuration module 701 is configured to configure the configuration information of the time series prediction model for the first communication device, wherein the time series prediction model is used to predict time series related information; the receiving module 702 is used to The target prediction information reported by the first communication device is received, wherein the target prediction information is the prediction information obtained by the first communication device using the time series prediction model to predict the first information.
  • the receiving module 702 is further configured to receive feedback information sent by the first communication device, where the feedback information is used to indicate that the first communication device supports the configuration information, Or the feedback information is used to indicate that the first communication device does not support the configuration information.
  • the receiving module 702 is further configured to receive the recommended configuration of the time series prediction model sent by the first communication device.
  • the configuration module 701 configures the configuration information of the time series prediction model for the first communication device, including:
  • the second information includes at least one of the following:
  • Performance requirement information wherein the performance requirement information includes at least one of the following: prediction accuracy requirement information, processing delay requirement information, and calculation delay requirement information.
  • the receiving module 702 is further configured to receive model configuration request information sent by the first communication device.
  • the apparatus for obtaining prediction information provided by the embodiments of the present application can implement the various processes implemented by the second communication device or the network side device in the method embodiments shown in Figures 2 to 5, and achieve the same technical effect. repeat.
  • this embodiment of the present application also provides a communication device 800, including a processor 801 and a memory 802, and the memory 802 stores programs or instructions that can run on the processor 801, for example
  • the communication device 800 is the first communication device, when the program or instruction is executed by the processor 801, each step of the above-mentioned model construction method embodiment can be realized, and the same technical effect can be achieved.
  • the communication device 800 is the second communication device, when the program or instruction is executed by the processor 801, the various steps of the method for obtaining the above-mentioned prediction information can be implemented, and the same technical effect can be achieved. To avoid repetition, details are not repeated here. .
  • the embodiment of the present application also provides a terminal, including a processor and a communication interface, the processor is used to implement the steps of the above-mentioned model construction method embodiment, and the communication interface is used to communicate with an external communication device.
  • This terminal embodiment corresponds to the above-mentioned first communication device side method embodiment, and each implementation process and implementation mode of the above-mentioned method embodiment can be applied to this terminal embodiment, and can achieve the same technical effect.
  • FIG. 9 is a schematic diagram of a hardware structure of a terminal implementing an embodiment of the present application.
  • the terminal 900 includes, but is not limited to: a radio frequency unit 901, a network module 902, an audio output unit 903, an input unit 904, a sensor 905, a display unit 906, a user input unit 907, an interface unit 908, a memory 909, and a processor 910. At least some parts.
  • the terminal 900 can also include a power supply (such as a battery) for supplying power to various components, and the power supply can be logically connected to the processor 910 through the power management system, so as to manage charging, discharging, and power consumption through the power management system. Management and other functions.
  • a power supply such as a battery
  • the terminal structure shown in FIG. 9 does not constitute a limitation on the terminal, and the terminal may include more or fewer components than shown in the figure, or combine some components, or arrange different components, which will not be repeated here.
  • the input unit 904 may include a graphics processor unit (Graphics Processing Unit, GPU) 9041 and a microphone 9042, and the graphics processor 9041 is used by the image capture device in the video capture mode or the image capture mode (such as a camera) to process the image data of still pictures or videos.
  • the display unit 906 may include a display panel 9061, and the display panel 9061 may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like.
  • the user input unit 907 includes at least one of a touch panel 9071 and other input devices 9072 .
  • the touch panel 9071 is also called a touch screen.
  • the touch panel 9071 may include two parts, a touch detection device and a touch controller.
  • Other input devices 9072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, switch buttons, etc.), trackballs, mice, and joysticks, which will not be repeated here.
  • the radio frequency unit 901 may transmit the downlink data from the network side device to the processor 910 for processing after receiving the downlink data; in addition, the radio frequency unit 901 may send the uplink data to the network side device.
  • the radio frequency unit 901 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 909 can be used to store software programs or instructions as well as various data.
  • the memory 909 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required by at least one function (such as a sound playing function, image playback function, etc.), etc.
  • memory 909 may include volatile memory or nonvolatile memory, or, memory 909 may include both volatile and nonvolatile memory.
  • the non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electronically programmable Erase Programmable Read-Only Memory (Electrically EPROM, EEPROM) or Flash.
  • ROM Read-Only Memory
  • PROM programmable read-only memory
  • Erasable PROM Erasable PROM
  • EPROM erasable programmable read-only memory
  • Electrical EPROM Electrical EPROM
  • EEPROM electronically programmable Erase Programmable Read-Only Memory
  • Volatile memory can be random access memory (Random Access Memory, RAM), static random access memory (Static RAM, SRAM), dynamic random access memory (Dynamic RAM, DRAM), synchronous dynamic random access memory (Synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (Double Data Rate SDRAM, DDRSDRAM), enhanced synchronous dynamic random access memory (Enhanced SDRAM, ESDRAM), synchronous connection dynamic random access memory (Synch link DRAM , SLDRAM) and Direct Memory Bus Random Access Memory (Direct Rambus RAM, DRRAM).
  • RAM Random Access Memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • DRAM synchronous dynamic random access memory
  • SDRAM double data rate synchronous dynamic random access memory
  • Double Data Rate SDRAM Double Data Rate SDRAM
  • DDRSDRAM double data rate synchronous dynamic random access memory
  • Enhanced SDRAM, ESDRAM enhanced synchronous dynamic random access memory
  • Synch link DRAM , SLDRAM
  • Direct Memory Bus Random Access Memory Direct Rambus
  • the processor 910 may include one or more processing units; optionally, the processor 910 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to the operating system, user interface, and application programs, etc., Modem processors mainly process wireless communication signals, such as baseband processors. It can be understood that the foregoing modem processor may not be integrated into the processor 910 .
  • the radio frequency unit 901 is configured to obtain configuration information of a time series prediction model from the second communication device, wherein the time series prediction model is used to predict time series related information;
  • the processor 910 is configured to apply the configuration information to configure the time series prediction model.
  • the embodiment of the present application also provides a network side device, including a processor and a communication interface, the processor is used to implement the various processes of the above embodiments of the method for obtaining prediction information, and the communication interface is used to communicate with an external communication device.
  • the network-side device embodiment corresponds to the above-mentioned second communication device method embodiment, and each implementation process and implementation mode of the above-mentioned method embodiment can be applied to this network-side device embodiment, and can achieve the same technical effect.
  • the embodiment of the present application also provides a network side device.
  • the network side device 1000 includes: an antenna 1001 , a radio frequency device 1002 , a baseband device 1003 , a processor 1004 and a memory 1005 .
  • the antenna 1001 is connected to the radio frequency device 1002 .
  • the radio frequency device 1002 receives information through the antenna 1001, and sends the received information to the baseband device 1003 for processing.
  • the baseband device 1003 processes the information to be sent and sends it to the radio frequency device 1002
  • the radio frequency device 1002 processes the received information and sends it out through the antenna 1001 .
  • the method performed by the network side device in the above embodiments may be implemented in the baseband device 1003, where the baseband device 1003 includes a baseband processor.
  • the baseband device 1003 may include at least one baseband board, on which a plurality of chips are arranged, as shown in FIG.
  • the program executes the network device operations shown in the above method embodiments.
  • the network side device may also include a network interface 1006, such as a common public radio interface (common public radio interface, CPRI).
  • a network interface 1006 such as a common public radio interface (common public radio interface, CPRI).
  • the network side device 1000 in the embodiment of the present application further includes: instructions or programs stored in the memory 1005 and executable on the processor 1004, and the processor 1004 calls the instructions or programs in the memory 1005 to execute the The method of module execution achieves the same technical effect, so in order to avoid repetition, it is not repeated here.
  • the embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the various processes of the above-mentioned model construction method embodiment are realized, or the above-mentioned Each process of the embodiment of the method for obtaining prediction information can achieve the same technical effect, and will not be repeated here to avoid repetition.
  • the processor is the processor in the terminal described in the foregoing embodiments.
  • the readable storage medium includes a computer-readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk, and the like.
  • the embodiment of the present application further provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the above-mentioned model construction method embodiment
  • the chip includes a processor and a communication interface
  • the communication interface is coupled to the processor
  • the processor is used to run programs or instructions to implement the above-mentioned model construction method embodiment
  • the chip mentioned in the embodiment of the present application may also be called a system-on-chip, a system-on-chip, a system-on-a-chip, or a system-on-a-chip.
  • the embodiment of the present application further provides a computer program/program product, 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 realize the implementation of the above-mentioned model construction method
  • the embodiment of the present application also provides a model configuration system, including: a first communication device and a second communication device, the first communication device can be used to execute the steps of the above-mentioned model construction method, and the network side The device can be used to execute the steps of the method for obtaining prediction information as described above.
  • the term “comprising”, “comprising” or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or device. Without further limitations, an element defined by the phrase “comprising a " does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising that element.
  • the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved. Functions are performed, for example, 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.
  • the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation.
  • the technical solution of the present application can be embodied in the form of computer software products, which are stored in a storage medium (such as ROM/RAM, magnetic disk, etc.) , CD-ROM), including several instructions to make a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of the present application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种模型的构建方法、装置及通信设备,属于无线通信领域。所述构建方法包括:第一通信设备从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息(S210);所述第一通信设备应用所述配置信息构建所述时间序列预测模型(S212)。

Description

模型的构建方法、装置及通信设备
相关申请的交叉引用
本申请要求在2021年12月30日提交的中国专利申请第202111662864.6号的优先权,该中国专利申请的全部内容通过引用包含于此。
技术领域
本申请属于无线通信技术领域,具体涉及一种模型的构建方法、装置及通信设备。
背景技术
由于移动设备和移动流量的快速增长以及大量应用场景的出现,未来的无线通信网络可能会具有高速率、低时延以及移动性增强等多种指标需求。例如,对于移动性增强,在非陆地网络(Non Terrestrial Networks,NTN)或高铁等多种高移动性场景,由于周围散射环境的快速变化,无线信道的相干时间被严重缩短,也就是说无线通信设备和网络侧基站对无线信道的探测结果可能迅速失效,所以必须进行更频繁的无线信道探测以维持良好的通信性能,这对网络的负载和能耗带来的巨大挑战。
时间序列预测是一种终端及网络侧设备基于对过去一段时间通信测量量的观察,预测未来的通信测量量的方法,本质上是基于历史测量量的数据挖掘他们之间时间相关性。然而,由于在实际应用中,终端的类型多种多样,能力各不相同,需求也可能不相同,因此,如何构建用于预测时间序列相关的信息的模型以使得通信双方能够理解一致是目前需要解决的技术问题。
发明内容
本申请实施例提供一种模型的构建方法、装置及通信设备,能够解决如何构建用于预测时间序列相关的信息的模型以使得通信双方能够理解一致的 问题。
第一方面,提供了一种模型的构建方法,包括:第一通信设备从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;所述第一通信设备应用所述配置信息构建所述时间序列预测模型。
第二方面,提供了一种模型的构建装置,包括:获取模块,用于从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;应用模块,用于应用所述配置信息构建所述时间序列预测模型。
第三方面,提供了一种预测信息的获取方法,包括:第二通信设备为第一通信设备配置时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;所述第二通信设备接收所述第一通信设备上报的目标预测信息,其中,所述目标预测信息为所述第一通信设备使用所述时间序列预测模型对第一信息进行预测得到的预测信息。
第四方面,提供了一种预测信息的获取装置,包括:配置模块,用于第二通信设备为第一通信设备配置时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;接收模块,用于接收所述第一通信设备上报的目标预测信息,其中,所述目标预测信息为所述第一通信设备使用所述时间序列预测模型对第一信息进行预测得到的预测信息。
第五方面,提供了一种通信设备,该终端包括处理器和存储器,所述存储器存储可在所述处理器上运行的程序或指令,所述程序或指令被所述处理器执行时实现如第一方面所述的方法的步骤,或实现如第三方面所述的方法的步骤。
第六方面,提供了一种通信设备,包括处理器及通信接口,其中,所述处理器用于实现如第一方面所述的方法的步骤,或实现如第一方面所述的方法的步骤,所述通信接口用于与外部设备进行通信。
第七方面,提供了一种模型的配置系统,包括:第一通信设备及第二通信设备,所述第一通信设备可用于执行如第一方面所述的方法的步骤,所述第二通信设备可用于执行如第三方面所述的方法的步骤。
第八方面,提供了一种可读存储介质,所述可读存储介质上存储程序或指令,所述程序或指令被处理器执行时实现如第一方面所述的方法的步骤,或者实现如第三方面所述的方法的步骤。
第九方面,提供了一种芯片,所述芯片包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现如第一方面所述的方法的步骤,或实现如第三方面所述的方法的步骤。
第十方面,提供了一种计算机程序/程序产品,所述计算机程序/程序产品被存储在存储介质中,所述计算机程序/程序产品被至少一个处理器执行以实现如第一方面所述的方法的步骤,或实现如第三方面所述的方法的步骤。
在本申请实施例中,第一通信设备从第二通信设备获取用于预测时间序列相关的信息的时间序列预测模型的配置信息,并应用该配置信息构建时间序列预测模型,从而使得第一通信设备可以通过该时间序列预测模型对时间序列相关的信息进行预测,从而解决了如何配置用于预测时间序列相关的信息的模型以使得通信双方能够理解一致的问题。
附图说明
图1示出本申请实施例可应用的一种无线通信系统的框图;
图2示出本申请实施例提供的模型的构建方法的一种流程示意图;
图3示出本申请实施例提供的预测信息的获取方法的一种流程示意图;
图4示出本申请实施例提供的模型的构建方法的另一种流程示意图;
图5示出本申请实施例中一种训练窗及训练样本采样的示意图;
图6示出本申请实施例提供的模型的构建装置的一种结构示意图;
图7示出本申请实施例提供的预测信息的获取装置的一种结构示意图;
图8示出本申请实施例提供的一种通信设备的结构示意图;
图9示出本申请实施例提供的一种终端的硬件结构示意图;
图10示出本申请实施例提供的一种网络侧设备的硬件结构示意图。
具体实施方式
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员所获得的所有其他实施例,都属于本申请保护的范围。
本申请的说明书和权利要求书中的术语“第一”、“第二”等是用于区别类似的对象,而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换,以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施,且“第一”、“第二”所区别的对象通常为一类,并不限定对象的个数,例如第一对象可以是一个,也可以是多个。此外,说明书以及权利要求中“和/或”表示所连接对象的至少其中之一,字符“/”一般表示前后关联对象是一种“或”的关系。
值得指出的是,本申请实施例所描述的技术不限于长期演进型(Long Term Evolution,LTE)/LTE的演进(LTE-Advanced,LTE-A)系统,还可用于其他无线通信系统,诸如码分多址(Code Division Multiple Access,CDMA)、时分多址(Time Division Multiple Access,TDMA)、频分多址(Frequency Division Multiple Access,FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access,OFDMA)、单载波频分多址(Single-carrier Frequency Division Multiple Access,SC-FDMA)和其他系统。本申请实施例中的术语“系统”和“网络”常被可互换地使用,所描述的技术既可用于以上提及的系统和无线电技术,也可用于其他系统和无线电技术。以下描述出于示例目的描述了新空口(New Radio,NR)系统,并且在以下大部分描述中使用NR 术语,但是这些技术也可应用于NR系统应用以外的应用,如第6代(6 th Generation,6G)通信系统。
图1示出本申请实施例可应用的一种无线通信系统的框图。无线通信系统包括终端11和网络侧设备12。其中,终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)或称为笔记本电脑、个人数字助理(Personal Digital Assistant,PDA)、掌上电脑、上网本、超级移动个人计算机(ultra-mobile personal computer,UMPC)、移动上网装置(Mobile Internet Device,MID)、增强现实(augmented reality,AR)/虚拟现实(virtual reality,VR)设备、机器人、可穿戴式设备(Wearable Device)、车载设备(VUE)、行人终端(PUE)、智能家居(具有无线通信功能的家居设备,如冰箱、电视、洗衣机或者家具等)、游戏机、个人计算机(personal computer,PC)、柜员机或者自助机等终端侧设备,可穿戴式设备包括:智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装等。需要说明的是,在本申请实施例并不限定终端11的具体类型。网络侧设备12可以包括接入网设备和/或核心网设备,其中,接入网设备12也可以称为无线接入网设备、无线接入网(Radio Access Network,RAN)、无线接入网功能或无线接入网单元。接入网设备12可以包括基站、WLAN接入点或WiFi节点等,基站可被称为节点B、演进节点B(eNB)、接入点、基收发机站(Base Transceiver Station,BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set,BSS)、扩展服务集(Extended Service Set,ESS)、家用B节点、家用演进型B节点、发送接收点(Transmission Reception Point,TRP)或所述领域中其他某个合适的术语,只要达到相同的技术效果,所述基站不限于特定技术词汇,需要说明的是,在本申请实施例中仅以NR系统中的基站为例进行介绍,并不限定基站的具体类型。
下面结合附图,通过一些实施例及其应用场景对本申请实施例提供的模 型的配置方案进行说明进行详细地说明。
图2示出本申请实施例提供的模型的构建方法的一种流程示意图,如图2所示,该方法200主要包括以下步骤。
S210,第一通信设备从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息。
在本申请实施例中,第一通信设备和第二通信设备可以分别为无线通信中的两端,例如,第一通信设备为终端,第二通信设备为网络侧设备。或者,第一通信设备为远端(Remote)终端,第二通信设备中继(Relay)终端。或者,第一通信设备和第二通信设备也可以为其它通信设备,具体本申请实施例中不作限定。
其中,网络侧设备包括但不限于以下至少之一:
a)核心网节点,包括NWDAF、LMF等,或神经网络处理节点。
b)接入网节点,包括基站或新定义的神经网络处理节点。
在本申请实施例中,所述时间序列预测模型预测的第一信息为时间序列相关的信息,例如,第一信息可以为信道状态信息(Channel State Information,CSI)、第一通信设备的位置信息、第一通信设备波束信息等与时间序列相关的信息中的一项或多项。
在一个可能的实现方式中,第二通信设备可以主动向第一通信设备发送上述配置信息,例如,第二通信设备根据第一通信设备当前所处的应用场景,主动向第一通信设备发送上述配置信息,指示第一通信设备部署时间序列预测模型对第一信息进行预测。
在另一个可能的实现方式中,第二通信设备也可以是基于第一通信设备的请求,向第一通信设备发送上述配置信息。因此,在该可能的实现方式中,在第一通信设备从第二通信设备获取时间序列预测模型的配置信息之前,所述方法还包括:所述第一通信设备向所述第二通信设备发送模型配置请求信息。例如,第一通信设备根据其能力,向所述第二通信设备发送模型配置请 求信息,请求第二通信设备为其配置上述配置信息。
S212,所述第一通信设备应用所述配置信息构建所述时间序列预测模型。
在本申请实施例中,第一通信设备根据所述配置信息,构建时间序列预测模型,从而可以使用构建的时间序列预测模型对第一信息进行预测,以减少对通信测量量的测量。
在本申请实施例中,第一通信设备从第二通信设备获取用于预测时间序列相关的信息的时间序列预测模型的配置信息,并应用该配置信息构建时间序列预测模型,从而使得第一通信设备可以通过该时间序列预测模型对时间序列相关的信息进行预测。从而解决了如何构建用于预测时间序列相关的信息的模型以使得通信双方能够理解一致的问题。另外,由于第一通信设备使用时间序列预测模型进行预测得到的预测信息需要反馈给第二通信设备,因此,通过由第二通信设备配置时间序列预测模型的配置信息的方式,可以使得第二通信设备能够获知接收到的预测信息对应的时间序列预测模型的配置信息,便于第二通信设备使用所述预测信息。另外,如果第一通信设备为终端,由于各个终端的能力不一样,通过由第二通信设备配置时间序列预测模型的配置信息的方式,可以为各个终端配置合适的时间序列预测模型。
在一个可能的实现方式中,所述时间序列预测模型可以为高斯过程(Gaussian Process,GP)模型。
其中,GP模型是基于统计学习理论和贝叶斯理论发展起来的一种机器学习方法,能够处理高维度、小样本和非线性的复杂回归和分类问题,具有较强的泛化能力。与神经网络等方法相比,GP具有容易实现、超参数自适应获取、灵活非参数推断、可解释以及概率意义的输出等优点,具有解决未来无线通信系统复杂时间序列预测问题的潜力。
例如,假设一种标准线性模型:
y=x Tw+∈=f(x)+∈
其中,∈服从0均值,方差为
Figure PCTCN2022143672-appb-000001
高斯分布:
Figure PCTCN2022143672-appb-000002
暂时不考虑噪声,GP模型通过对有限样本<x,y>的观察,估计参数w,其中,x为GP模型的输入,y为GP模型标签。根据最大似然估计理论,似然概率表示如下:
Figure PCTCN2022143672-appb-000003
根据贝叶斯定理,需要定义关于参数w的先验信息(在观察x,y之前,对参数w的先验认知),假设w服从均值为0,协方矩阵差为∑ p的高斯分布:
w~N(0,∑ p);
GP模型是一种有监督机器学习方法,基于输入X和输出y的观测利用最大后验概率准则推断模型参数:
Figure PCTCN2022143672-appb-000004
其中,边际概率与参数w无关,是一个归一化的常数:
p(w|y,X)=∫p(y|X,w)p(w)dw;
进一步:
Figure PCTCN2022143672-appb-000005
其中:
Figure PCTCN2022143672-appb-000006
进一步得到参数w的最大后验概率:
Figure PCTCN2022143672-appb-000007
其中,
Figure PCTCN2022143672-appb-000008
GP模型考虑了所有w的可能性,对所有可能的线性模型组合取加权平 均,输出的后验分布同样服从高斯分布,将均值作为预测值:
Figure PCTCN2022143672-appb-000009
预测值:
Figure PCTCN2022143672-appb-000010
可以看出,预测值为测试输入x *乘以权重参数w的最大似然估计,即
Figure PCTCN2022143672-appb-000011
的最大似然估计(无噪声)。预测的不确定性(方差):
Figure PCTCN2022143672-appb-000012
因此,GP模型可以在有限观测样本(x,y)的情况下进行模型训练,并且当测试数据输入x *输入到GP模型,可以同时给出对标签f *和预测不确定性(方差)的估计。
上述给出了简略的线性无噪声GP模型的理论推导过程,对于非线性模型,GP的处理方式首先通过核函数(Kernel function)将低维空间下的非线性模型转换为高维空间的线性模型,即通过一组基函数Φ(·),将有限维空间的样本向量 x映射到高维空间,标准线性模型进一步拓展为:
f(x)=Φ(x) Tw;
该模型是x维度拓展后标准线性模型,处理方式与标准线性模型相同,即测试输入为x *的情况下的预测标签f *的服从如下条件分布:
Figure PCTCN2022143672-appb-000013
其中:
Figure PCTCN2022143672-appb-000014
Φ(x)是一个高维甚至无穷维的向量,高维向量的内积的计算复杂度较高,这里引入kernel近似高维向量的内积:
Figure PCTCN2022143672-appb-000015
其中,
Figure PCTCN2022143672-appb-000016
其中,常见的kernel包括:径向基核:
Figure PCTCN2022143672-appb-000017
以及有理二次核:
Figure PCTCN2022143672-appb-000018
等,其中d为欧式距离,α,l为超参数。由于无穷阶可导的圆滑特性,径向基核的应用最为广泛,并且相关理论证明径向基核可以拆分成两个无穷维向量的内积,能够实现有限维向量x向无穷维向量的映射。
1、对于noisy-free的情况:y=f(x)
测试样本输出服从高斯分布:
Figure PCTCN2022143672-appb-000019
因此,测试输出和训练输出的联合分布为:
Figure PCTCN2022143672-appb-000020
根据联合分布可以得到条件高斯概率分布:
Figure PCTCN2022143672-appb-000021
条件高斯概率分布的均值即为预测值的最大后验概率估计,方差表示预测的不确定性(方差):
Figure PCTCN2022143672-appb-000022
Figure PCTCN2022143672-appb-000023
K(X *,X)=K(X,X *) T
Figure PCTCN2022143672-appb-000024
Figure PCTCN2022143672-appb-000025
X=[x 1,x 2,x 3,..,x n],则
Figure PCTCN2022143672-appb-000026
Figure PCTCN2022143672-appb-000027
其中,n *和n分别为测试集和训练集的样本数。
2、对于noisy的情况:y=f(x)+n
Figure PCTCN2022143672-appb-000028
Figure PCTCN2022143672-appb-000029
Figure PCTCN2022143672-appb-000030
相对于把noisy-free情况下的K(X,X)换成
Figure PCTCN2022143672-appb-000031
进一步可以写成一种更紧凑的形式:
Figure PCTCN2022143672-appb-000032
Figure PCTCN2022143672-appb-000033
3、核函数参数的优化:
根据训练窗数据,最大化边缘概率估计模型的最优超参数:
Figure PCTCN2022143672-appb-000034
其中:
Figure PCTCN2022143672-appb-000035
因此,在一个可能的实现方式中,所述配置信息包括以下至少之一:
(1)核函数的第一配置信息。
其中,所述第一配置信息可以包括:
核函数的数量。例如,该时间序列预测模型使用的核函数的数量。
核函数的超参数。例如,该时间序列预测模型使用的核函数的超参数。
核函数的类型。例如,该时间序列预测模型使用的哪几个核函数,如平方指数核、二次有理核、常数核以及其他自定义核函数等。
(2)第一任务标识,其中,所述第一任务标识用于指示所述 第一配置信息相关联的任务。
例如,该第一任务标识指示与(1)中的核函数配置相关联的任务,时间序列预测模型应用于预测第一信息,第一信息可以包括CSI、位置信息、波束信息等与时间序列相关信息。
(3)优化器选择信息。其中,优化器用于优化核函数中的超参数。
例如,优化器选择信息可以包括以下之一:
a)梯度下降优化法的第一配置参数,其中,所述第一配置参数包括:初始值、步长、和终止条件。即优化器采用梯度下降优化法,优化器选择信息中包括梯度下降优化法中的初始值设置、步长设置以及终止条件等。
b)网格搜索优化法的第二配置参数,其中,所述第二配置参数包括:不同超参数网格搜索的上界、不同超参数网格搜索的下界、和不同超参数网络搜索的步长。即优化器采用网格搜索法,优化器选择信息中包括:不同超参数网格搜索的上界;不同超参数网格搜索的下界;不同超参数网格搜索的步长等信息。
c)优化器的运行次数。例如,时间序列预测模型在训练过程中,上述a)或b)所述优化器的执行次数。
(4)第一模型训练配置。
其中,所述第一模型训练配置包括以下至少之一:
a)第一训练窗的长度。例如,训练数据中包含N个时间单位的CSI,则第一训练窗长度为N个时间单位。
b)模型训练的输入。例如,模型训练的输入的时间单位编号{1,2,…,N}。
c)模型训练的标签。例如,与模型训练输入的时间单位编号对应的CSI。
d)第一训练窗样本采样的时间间隔。即第一训练窗内间隔多长时间采集一次样本。
(5)第一模型预测配置。
其中,所述第一模型预测配置可以包括以下至少之一:
a)第一预测窗的长度;例如,基于N个时间单位的CSI预测未来M个时间单位的CSI,则第一预测窗的长度为M个时间单位。
b)预测的输入;例如,基于N个时间单位的训练数据预测未来M个时间单位的CSI,则模型预测的输入为M个时间单位CSI对应的时间单位编号,例如,{N+1,…,N+M}。
c)预测的输出。
其中,时间序列预测模型的预测的输出可以包括:与模型预测输入的时间单位编号相对应的标签信息,例如,{N+1,…,N+M};模型预测标签的误差信息,例如,时间序列预测模型输出的预测方差和/或所述时间序列预测模型输出的预测误差。
d)第一预测窗样本采样的时间间隔。即第一预测窗中两个样本之间间隔的时间。
(6)所述时间序列预测模型的生命周期。
其中,所述时间序列预测模型的生命周期包括但不限于以下之一:生效时间、失效时间、以及运行时间。
(7)所述时间序列预测模型的计算模式配置。
其中,所述计算模式配置包括以下之一:
a)串行计算;即GP模型训练在同一计算单元进行,例如,在CPU某一核或线程内进行。
b)并行计算,即GP模型训练在多个计算单元同时进行,如在CPU多个核或、线程内进行。
需要说明的是,本申请实施例中的时间单位是预先设置的单位时间,例 如,时间单位可以包括以下至少之一:参考信号周期、预测周期、符号(OFDM符号)、子帧、无线帧、毫秒、秒等。具体本申请实施例中不作限定。
在一个可能的实现方式中,上述(1)中核函数的第一配置信息指示的所述核函数为核函数列表中的至少一个核函数,其中,所述第一通信设备和所述第二通信设备中预先配置有所述核函数列表。其中,核函数列表中可以包括一个或多个核函数的标识,通过该标识可以唯一确定一个核函数。
在一个可能的实现方式中,上述(2)中的所述任务标识为任务列表中的至少一个任务,其中,所述第一通信设备和所述第二通信设备中预先配置有所述任务列表。也就是说,第一通信设备事先与第二通信设备配置相同的时间序列预测相关的任务列表,任务可由对应的任务标识唯一确定。
在一个可能的实现方式中,上述(3)中的所述优化器选择信息中包括目标优化器标识,所述目标优化器标识为优化器列表中的至少一个优化器标识,所述优化器列表中包括优化器标识与参数配置的对应关系,所述第一通信设备和所述第二通信设备中预先配置有所述优化器列表。也就是说,第一通信设备事先与第二通信设备配置相同的优化器选择及具体参数配置列表,优化器选择及配置信息可由列表标识唯一确定。
在一个可能的实现方式中,上述(4)中的所述模型训练配置还可以包括:
a)对训练输入的第一处理方式,其中,所述第一处理方式包括:线性放缩处理或非线性放缩处理;
b)对标签的第二处理方式,其中,所述第二处理方式包括:线性归一化处理(例如,scale*X)或均值归一化处理(例如,log2(X))。
在一个可能的实现方式中,所述计算模式配置可以由所述第一通信设备根据其计算能力和存储能力确定,也可以由第二通信设备向第一通信设备动态配置。
在一个可能的实现方式中,在第一通信设备从第二通信设备获取时间序列预测模型的配置信息之后,该方法还可以包括:所述第一通信设备向所述 第二通信设备发送反馈信息,其中,所述反馈信息用于指示所述第一通信设备支持所述配置信息、或者所述反馈信息用于指示所述第一通信设备不支持所述配置信息。在该可能的实现方式中,第一通信设备在从第二通信设备获取到所述配置信息之后,可以向第二通信设备反馈所述第一通信设备是否支持所述配置信息,从而使得第二通信设备可以获取第一通信设备是否可以配置与所述配置信息对应的时间序列预测模型。
在一个可能的实现方式中,所述反馈信息包括但不限于以下至少之一:
(1)第二指示信息,所述第二指示信息用于指示所述第一通信设备所支持的核函数,其中,所述核函数为所述配置信息中配置的一个或多个核函数;也就是说,第二指示信息用于指示第一通信设备支持配置信息配置的核函数中的哪一种或几种。
(2)第三指示信息,其中,所述第三指示信息用于指示所述第一通信设备具有任务预测的需求,或者,所述第三指示信息用于指示所述第一通信设备不具有所述任务预测的需求,所述任务为所述配置信息配置的任务;也就是说,第三指示信息用于指示第一通信设备是否具有配置信息配置的相关任务的预测需求。
(3)第四指示信息,其中,所述第四指示信息用于指示所述第一通信设备支持目标优化器,或者,所述第四指示信息用于指示所述第一通信设备不支持目标优化器,所述目标优化器为所述配置信息配置的优化器;也就是说,第四指示信息用于指示第一通信设备是否支持配置信息配置的优化器。
(4)第五指示信息,其中,所述第五指示信息用于指示所述第一通信设备支持所述配置信息中的模型训练配置,或者,所述第五指示信息用于指示所述第一通信设备不支持所述配置信息中的模型训练配置;也就是说,第五指示信息指示第一通信设备是否支持配置信息配置的模型训练配置。例如,第一通信设备可以考虑到其存储能 力和/或计算能力有限,不支持较长的训窗尺寸,如果配置信息配置的第一训练窗的长度超过了第一通信设备的存储能力和/或计算能力,则第一通信设备可以反馈不支持所述配置信息中的模型训练配置。
(5)第六指示信息,其中,所述第六指示信息用于指示所述第一通信设备支持所述配置信息中的模型预测配置,或者,所述第六指示信息用于指示所述第一通信设备不支持所述配置信息中的模型预测配置。也就是说,第六指示信息用于指示第一通信设备是否支持所述配置信息中的模型预测配置,例如,是否支持所述配置信息配置的预测窗的长度等。
在一个可能的实现方式中,为了使第二通信设备配置的配置信息能够更加符合第一通信设备的需求,在第一通信设备从第二通信设备获取时间序列预测模型的配置信息之前,所述方法还可以包括:所述第一通信设备根据第二信息,向所述第二通信设备推荐所述时间序列预测模型的推荐配置。通过该可能的实现方式,可以使得第二通信设备为第一通信设备配置的配置信息能更适应于第一通信设备。
在一个可能的实现方式中,所述推荐配置包括以下至少之一:
(1)推荐的核函数;即第一通信设备推荐使用的核函数。
(2)推荐的模型训练配置;即第一通信设备推荐使用的模型训练配置,例如,第一训练窗的长度等。
(3)推荐的模型预测配置。即第一通信设备推荐使用的模型预测配置,例如,预测窗的长度等。
在一个可能的实现方式中,所述第二信息包括以下至少之一:
(1)待预测的第一信息的统计信息,例如,第一信息的均值和/或方差;
(2)第一信息的估计误差;基于已知的信息预测未知信息,因此,第一信息的估计误差是指对已知信息(历史信息)的估计误差。 例如,对于CSI预测,过去N个时间单位的CSI是终端基于导频(SRS)估计的,估计是存在误差的;
(3)所述估计误差的统计信息;例如,第一信息的估计误差的均值和/或方差。
(4)模型预测误差;
(5)所述模型预测误差的统计信息;例如,模型预测误差的均值和/或方差。
(6)所述第一通信设备的移动性信息;例如,第一通信设备的移动速度、波束切换信息、小区切换信息等中的至少一项。
(7)噪声的统计信息;例如,信噪比、信干噪比等。
(8)性能需求信息,其中,所述性能需求信息包括以下至少一项:预测精度的需求信息、处理时延的需求信息、计算时延的需求信息。
在一个可能的实现方式中,在所述第一通信设备应用所述配置信息配置所述时间序列预测模型之后,可以基于上述的模型训练配置对时间序列预测模型进行训练,在训练之后,基于上述的模型预测配置使用所述时间序列预测模型对第一信息进行预测。其中,在使用时间序列预测模型进行预测时,预测输入、输出的处理方式与训练时的训练输入、输出的处理方式相同。
在一个可能的实现方式中,在第一通信设备应用所述配置信息配置所述时间序列预测模型之后,所述方法还可以包括:所述第一通信设备向所述第二通信设备上报目标预测信息,其中,所述目标预测信息为使用所述时间序列预测模型对所述第一信息进行预测得到的预测信息。即第一通信设备向第二通信设备上报预测的第一信息。例如,第一通信设备根据已测量得到的N个时间单位的CSI信息,通过时间序列预测模型预测未来M个时间单位的CSI信息,并上报预测得到的M个时间单位的CSI信息。
在一个可能的实现方式中,所述目标预测信息包括以下至少之一:
(1)预测得到的第一信息;
(2)预测所述第一信息对应的预测误差;
(3)第二任务标识;该第二任务标识用于指示所述第一信息的具体类别,例如,为CSI或波束信息等。在具体应用中,第二任务标识可以是上述第一任务标识,或者,也可以是上述第一任务标识中的一个。另外,第一通信设备可能同时有多种预测任务,第二通信设备基于每个预测任务配置一个配置信息,第一通信设备在上报目标预测信息中携带第二任务标识可以指示当前上报的目标预测信息所对应的预测任务。
(4)第二训练窗的配置相关信息;即预测第一信息对应的时间序列预测模型的训练窗配置的相关信息,该第二训练窗与上述配置信息中配置的训练窗配置对应的训练窗可以相同,也可能是上述配置信息中配置的训练窗配置对应的训练窗中的一个。
(5)第二预测窗的相关信息。
在一个可能的实现方式中,第二训练窗的配置相关信息可以包括以下至少之一:
a)第二训练窗的长度;例如,训练数据中包含N个时间单位的CSI,N为训练窗长度。
b)模型训练的输入;如时间单位编号{1,2,…,N}。
c)第一时间戳信息,其中,所述第一时间戳信息包括以下至少之一:第二训练窗的编号、第二训练窗的开始时间(即开始时刻)、第二训练窗的结束时间(即结束时刻)、第二训练窗样本采样的时间间隔(即第二训练窗中两个样本之间间隔的时间)。例如,如第二训练窗的CSI为第0、2、4、6、8时间单位的CSI,则第二训练窗样本采样的时间间隔为2个时间单位。
在一个可能的实现方式中,所述第二预测窗的配置相关信息包括以下至 少之一:
a)第二预测窗的长度;
b)模型预测的输入;
c)第二时间戳信息,其中,所述第二时间戳信息包括以下至少之一:第二预测窗的编号、第二预测窗的开始时间、第二预测窗的结束时间、第二预测窗样本采样的时间间隔。
在一个可能的实现方式中,所述时间序列预测模型可以基于神经网络的时间序列预测模型。
在上述可能的实现方式中,可选地,所述配置信息包括以下至少之一:
(1)所述时间序列预测模型的结构信息;
可选地,时间序列预测模型的结构信息可以包括:
a)神经网络的类型,例如,卷积神经网络、全连接神经网络、Transformer等。
b)神经网络的超参数信息,其中,该超参数信息可以包括:
i.输入输出的维度
ii.神经网络的层数及各层神经元数量
iii.激活函数
iv.关键模块的数量,例如,若采用Transformer模型,则可以包括Transformer的编码(encoder)模块的数量和/或解码(decoder)模块的数量。
(2)所述时间序列预测模型的权值;例如,神经网络中各个节点的权值、偏置等;
(3)所述时间序列预测模型的配置,其中,所述配置包括以下至少之一:优化器、损失函数;例如,通过编译(compile)模型的一些信息,如优化器、损失函数等。
(4)所述时间序列预测模型的优化器的状态信息。
在一个可能的实现方式中,第一通信设备从第二通信设备获取时间序列预测模型的配置信息,包括以下之一:
(1)获取第二通信设备非周期性配置的所述配置信息;例如,第一通信设备向第二通信设备发送请求,第二通信设备向第一通信设备配置所述配置信息。或者,第二通信设备突发性的为第一通信设备配置所述配置信息;
(2)接收所述第二通信设备周期性配置的所述配置信息;例如,第二通信设备以设定的周期,周期性的向第一通信设备配置的所述配置信息。
(3)所述第一通信设备与所述第二通信设备约定所述配置信息,即静态配置所述配置信息。
(4)所述第一通信设备确定所述配置信息,并向所述第二通信设备上报所述配置信息,接收到所述第二通信设备的确认信息后,获取所述配置信息。
图3示出本申请实施例提供的预测信息的获取方法的一种流程示意图,如图3所示,该方法300主要包括:
S310,第二通信设备为第一通信设备配置时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测第一信息,所述第一信息为时间序列相关的信息。
其中,所述配置信息与方法200中的配置信息相同,具体参见方法200中的描述。
S312,所述第二通信设备接收所述第一通信设备上报的目标预测信息,其中,所述目标预测信息为所述第一通信设备使用所述时间序列预测模型对所述第一信息进行预测得到的预测信息。
方法300是与方法200对应的第二通信设备侧的执行方法,具有与方法200相同或相应的可能实现方式,具体可以参见方法200中的描述。下面只对方法300中涉及的部分可能实现方式进行说明。
在一个可能的实现方式中,所述配置信息包括以下至少之一:
(1)核函数的第一配置信息,其中,所述第一配置信息包括: 核函数的数量、核函数的超参数以及核函数的类型;
(2)任务标识,其中,所述任务标识用于指示所述第一配置信息相关联的任务;
(3)优化器选择信息;
(4)模型训练配置,其中,所述模型训练配置包括以下至少之一:第一训练窗的长度、模型训练的输入、模型训练的标签、第一训练窗样本采样的时间间隔;
(5)模型预测配置,其中,所述模型预测配置包括以下至少之一:预测窗的长度、预测的输入、预测的输出、预测窗样本采样的时间间隔;
(6)所述时间序列预测模型的生命周期;
(7)所述时间序列预测模型的计算模式配置,其中,所述计算模式配置包括:串行计算或并行计算。
在一个可能的实现方式中,所述优化器选择信息包括以下之一:
(1)梯度下降优化法的第一配置参数,其中,所述第一配置参数包括:初始值、步长、和终止条件;
(2)网格搜索优化法的第二配置参数,其中,所述第二配置参数饿:不同超参数网格搜索的上界、不同超参数网格搜索的下界、和不同超参数网络搜索的步长。
在一个可能的实现方式中,所述预测的输出包括:
(1)与所述预测输入的时间单位编号对应的标签信息;以及
(2)所述时间序列预测模型输出的预测方差和/或所述时间序列预测模型输出的预测误差。
在一个可能的实现方式中,所述时间序列预测模型的生命周期包括:生效时间、失效时间、以及运行时间。
在一个可能的实现方式中,在第二通信设备为第一通信设备配置时间序 列预测模型的配置信息之后,所述方法还包括:
所述第二通信设备接收所述第一通信设备发送的反馈信息,其中,所述反馈信息用于指示所述第一通信设备支持所述配置信息、或者所述反馈信息用于指示所述第一通信设备不支持所述配置信息。
在一个可能的实现方式中,所述反馈信息包括以下至少之一:
(1)第二指示信息,所述第二指示信息用于指示所述第一通信设备所支持的核函数,其中,所述核函数为所述配置信息中配置的一个或多个核函数;
(2)第三指示信息,其中,所述第三指示信息用于指示所述第一通信设备具有任务预测的需求,或者,所述第三指示信息用于指示所述第一通信设备不具有所述任务预测的需求,所述任务为所述配置信息配置的任务;
(3)第四指示信息,其中,所述第四指示信息用于指示所述第一通信设备支持目标优化器,或者,所述第四指示信息用于指示所述第一通信设备不支持目标优化器,所述目标优化器为所述配置信息配置的优化器;
(4)第五指示信息,其中,所述第五指示信息用于指示所述第一通信设备支持所述配置信息中的模型训练配置,或者,所述第五指示信息用于指示所述第一通信设备不支持所述配置信息中的模型训练配置;
(5)第六指示信息,其中,所述第六指示信息用于指示所述第一通信设备支持所述配置信息中的模型预测配置,或者,所述第六指示信息用于指示所述第一通信设备不支持所述配置信息中的模型预测配置。
在一个可能的实现方式中,所述核函数为核函数列表中的至少一个核函数,其中,所述第一通信设备和所述第二通信设备中预先配置有所述核函数 列表。
在一个可能的实现方式中,所述任务标识为任务列表中的至少一个任务,其中,所述第一通信设备和所述第二通信设备中预先配置有所述任务列表。
在一个可能的实现方式中,所述优化器选择信息中包括目标优化器标识,所述目标优化器标识为优化器列表中的至少一个优化器标识,所述优化器列表中包括优化器标识与参数配置的对应关系,所述第一通信设备和所述第二通信设备中预先配置有所述优化器列表。
在一个可能的实现方式中,所述模型训练配置还包括:
(1)对训练输入的第一处理方式,其中,所述第一处理方式包括:线性放缩处理或非线性放缩处理;
(2)对标签的第二处理方式,其中,所述第二处理方式包括:线性归一化处理或均值归一化处理。
在一个可能的实现方式中,所述计算模式配置由所述第一通信设备根据其计算能力和存储能力确定。
在一个可能的实现方式中,在第二通信设备为第一通信设备配置时间序列预测模型的配置信息之前,所述方法还包括:所述第二通信设备接收所述第一通信设备发送的所述时间序列预测模型的推荐配置。
在一个可能的实现方式中,所述推荐配置包括以下至少之一:
推荐的核函数;
推荐的模型训练配置;
推荐的模型预测配置。
在一个可能的实现方式中,第二通信设备为第一通信设备配置时间序列预测模型的配置信息,包括:
所述第二通信设备接收所述第一通信设备上报的第二信息;
基于所述第二信息,为所述第一通信设备配置时间序列预测模型的配置信息;
其中,所述第二信息包括以下至少之一:
第一信息的统计信息;
第一信息的估计误差;
所述估计误差的统计信息;
模型预测误差;
所述模型预测误差的统计信息;
所述第一通信设备的移动性信息;
噪声的统计信息;
性能需求信息,其中,所述性能需求信息包括以下至少一项:预测精度的需求信息、处理时延的需求信息、计算时延的需求信息。
在一个可能的实现方式中,所述目标预测信息包括以下至少之一:
预测得到的第一信息;
预测所述第一信息对应的预测误差;
任务标识;
第二训练窗配置的相关信息;
第二预测窗的相关信息。
在一个可能的实现方式中,所述第二训练窗配置的相关信息包括以下至少之一:
第二训练窗的长度;
模型训练的输入;
第一时间戳信息,其中,所述第一时间戳信息包括以下至少之一:第二训练窗的编号、第二训练窗的开始时间、第二训练窗的结束时间、第二训练窗样本采样的时间间隔。
在一个可能的实现方式中,所述第二预测窗的相关信息包括以下至少之一:
第二预测窗的长度;
模型预测的输入;
第二时间戳信息,其中,所述第二时间戳信息包括以下至少之一:第二预测窗的编号、第二预测窗的开始时间、第二预测窗的结束时间、第二预测窗样本采样的时间间隔。
在一个可能的实现方式中,在第二通信设备为第一通信设备配置时间序列预测模型的配置信息之前,所述方法还包括:接收所述第一通信设备发送的模型配置请求信息。
在一个可能的实现方式中,所述时间序列预测模型包括:基于神经网络的时间序列预测模型;其中,所述配置信息包括以下至少之一:
所述时间序列预测模型的结构信息;
所述时间序列预测模型的权值;
所述时间序列预测模型的配置,其中,所述配置包括以下至少之一:优化器、损失函数;
所述时间序列预测模型的优化器的状态信息。
在一个可能的实现方式中,所述第二通信设备为第一通信设备配置时间序列预测模型的配置信息,包括以下之一:
(1)所述第二通信设备非周期性为第一通信设备配置时间序列预测模型的配置信息;
(2)所述第二通信设备周期性为第一通信设备配置时间序列预测模型的配置信息;
(3)所述第二通信设备与所述第一通信设备约定所述配置信息;
(4)所述第二通信设备所述第二通信设备上报所述配置信息,向所述第一通信设备发送确认信息。
下面以终端对CSI进行预测为例,对本申请实施例提供的技术方案进行说明。
图4示出本申请实施例提供的模型的构建方法的另一种流程图,如图3所示,该方法400主要包括以下步骤。
S401,终端向核心网设备请求GP模型配置信息。
其中,请求的GP模型配置信息包括:
a)核函数配置:径向基核函数、有理二次核函数两种核函数作为CSI预测的核函数:
Figure PCTCN2022143672-appb-000036
Figure PCTCN2022143672-appb-000037
b)任务标识:“1”用于指示CSI预测。
c)优化器选择:为网格搜索法,其中不同核函数的网格搜索范围均设置为:[-10,10],步长为1;
d)优化器运行次数:1次;
e)模型训练配置:训练窗长度为N个时间单位,即训练窗中历史CSI的数量为N,模型训练输入为{1,2,…,N},模型训练的标签为时间单位编号{1,2,…,N}对应的CSI,训练窗样本采样的时间间隔为2个时间单位;
f)模型预测配置信息:预测窗长度为M个时间单位,即基于历史N个时间单位的CSI预测未来M个时间单位的CSI,模型测试输入为时间单位编号{N+1,N+2,…,N+M},预测输出为时间单位编号{N+1,N+2,…,N+M}对应的CSI以及对应的预测方差。
以时间单位为1个时隙为例,如图5所示,训练窗的长度为6个时隙,训练窗样本采样间隔为2个时隙,预测窗的长度为1个时隙,预测窗采样间隔为2个时隙。
g)GP模型从第s个时间单位开始生效,第s’个时间单位失效.
S402,网络侧设备向终端配置GP模型相关信息。
S403,终端向网络侧设备上报GP模型预测信息。
其中,上报的预测信息可以包括:
a)预测的未来M个时间单位的CSI;
b)预测的M个时间单位的CSI对应的M个方差;
c)任务标识:“1”标识CSI预测;
d)训练窗配置相关信息,
e)预测窗配置信息,
其中,上报的训练窗配置相关信息可以包括:
i.训练窗长度为6个时间单位;
ii.模型训练输入为时间单位编号{1,2,…,6};
iii.时间戳信息。
其中,该时间戳信息可以包括:
1.训练窗编号为T,表示相比于开始训练训练窗的相对时间;
2.训练窗的开始时间为s,表示训练的开始时间;
3.训练窗的结束时间为s’,表示训练的结束时间;
4.训练窗样本采样的时间间隔为2,利用第{0,2,…,10,12}时间单位的CSI进行训练。
其中,预测窗配置信息可以包括:
i.预测窗长度为1,基于训练窗的6个时间单位的CSI预测未来1个时间单位的CSI;
ii.模型预测的输入,时间单位编号{7};
iii.时间戳信息。其中,该时间戳信息可以包括:
1.预测窗的编号为T’,表示相比于开始预测预测窗的相对时间;
2.预测窗的开始时间为p,表示预测的开始时间;
3.预测窗的结束时间为p’,表示预测的结束时间;
4.预测窗样本采样的时间间隔为2,预测第{14}个时间单位的CSI。
通过本申请实施例提供的上述技术方案中,提供了无线通信系统中将基于时间序列预测模型应用于无线通信关键指标预测时,交互的配置信息及流程,使得终端可以基于网络侧设备的配置部署时间序列预测模型,对与时间序列相关的参数进行预测。
本申请实施例提供的模型的构建方法,执行主体可以为模型的构建装置。本申请实施例中以模型的构建装置执行模型的构建方法为例,说明本申请实施例提供的模型的构建装置。
图6示出本申请实施例提供的模型的构建装置的一种结构示意图,如图6所示,该装置600主要包括:获取模块601和应用模块602。
在本申请实施例中,获取模块601,用于从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;应用模块602,用于应用所述配置信息构建所述时间序列预测模型。
在一个可能的实现方式中,所述装置还包括:第一发送模块,用于向所述第二通信设备发送反馈信息,其中,所述反馈信息用于指示所述第一通信设备支持所述配置信息、或者所述反馈信息用于指示所述第一通信设备不支持所述配置信息。
在一个可能的实现方式中,所述装置还包括:推荐模块,用于根据第二信息,向所述第二通信设备推荐所述时间序列预测模型的推荐配置。
在一个可能的实现方式中,所述装置还包括:上报模块,用于向所述第二通信设备上报目标预测信息,其中,所述目标预测信息为使用所述时间序列预测模型对所述第一信息进行预测得到的预测信息。
本申请实施例中的模型的构建装置可以是电子设备,例如具有操作系统的电子设备,也可以是电子设备中的部件,例如集成电路或芯片。该电子设备可以是终端,也可以为除终端之外的其他设备。示例性的,终端可以包括但不限于上述所列举的终端11的类型,其他设备可以为服务器、网络附属存储器(Network Attached Storage,NAS)等,本申请实施例不作具体限定。
本申请实施例提供的模型的构建装置能够实现图2至图5的方法实施例中第一通信设备或终端实现的各个过程,并达到相同的技术效果,为避免重复,这里不再赘述。
图7示出本申请实施例提供的预测信息的获取装置的一种结构示意图,如图7所示,该装置700主要包括:配置模块701和接收模块702。
在本申请实施例中,配置模块701,用于为第一通信设备配置时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;接收模块702,用于接收所述第一通信设备上报的目标预测信息,其中,所述目标预测信息为所述第一通信设备使用所述时间序列预测模型对第一信息进行预测得到的预测信息。
在一个可能的实现方式中,所述接收模块702,还用于接收所述第一通信设备发送的反馈信息,其中,所述反馈信息用于指示所述第一通信设备支持所述配置信息、或者所述反馈信息用于指示所述第一通信设备不支持所述配置信息。
在一个可能的实现方式中,所述接收模块702,还用于接收所述第一通信设备发送的所述时间序列预测模型的推荐配置。
在一个可能的实现方式中,所述配置模块701为第一通信设备配置时间序列预测模型的配置信息,包括:
接收所述第一通信设备上报的第二信息;
基于所述第二信息,为所述第一通信设备配置时间序列预测模型的配置信息;
其中,所述第二信息包括以下至少之一:
第一信息的统计信息;
第一信息的估计误差;
所述估计误差的统计信息;
模型预测误差;
所述模型预测误差的统计信息;
所述第一通信设备的移动性信息;
噪声的统计信息;
性能需求信息,其中,所述性能需求信息包括以下至少一项:预测精度的需求信息、处理时延的需求信息、计算时延的需求信息。
在一个可能的实现方式中,所述接收模块702,还用于接收所述第一通信设备发送的模型配置请求信息。
本申请实施例提供的预测信息的获取装置能够实现图2至图5的方法实施例中第二通信设备或网络侧设备实现的各个过程,并达到相同的技术效果,为避免重复,这里不再赘述。
可选的,如图8所示,本申请实施例还提供一种通信设备800,包括处理器801和存储器802,存储器802上存储有可在所述处理器801上运行的程序或指令,例如,该通信设备800为第一通信设备时,该程序或指令被处理器801执行时实现上述模型的构建方法实施例的各个步骤,且能达到相同的技术效果。该通信设备800为第二通信设备时,该程序或指令被处理器801执行时实现上述预测信息的获取方法实施例的各个步骤,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例还提供一种终端,包括处理器和通信接口,处理器用于实现上述模型的构建方法实施例的各个步骤,通信接口用于与外部通信设备进行通信。该终端实施例与上述第一通信设备侧方法实施例对应,上述方法实施例的各个实施过程和实现方式均可适用于该终端实施例中,且能达到相同的技术效果。具体地,图9为实现本申请实施例的一种终端的硬件结构示意图。
该终端900包括但不限于:射频单元901、网络模块902、音频输出单元903、输入单元904、传感器905、显示单元906、用户输入单元907、接口单元908、存储器909以及处理器910等中的至少部分部件。
本领域技术人员可以理解,终端900还可以包括给各个部件供电的电源(比如电池),电源可以通过电源管理系统与处理器910逻辑相连,从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。图9中示出的终端结构并不构成对终端的限定,终端可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置,在此不再赘述。
应理解的是,本申请实施例中,输入单元904可以包括图形处理器单元(Graphics Processing Unit,GPU)9041和麦克风9042,图形处理器9041对在视频捕获模式或图像捕获模式中由图像捕获装置(如摄像头)获得的静态图片或视频的图像数据进行处理。显示单元906可包括显示面板9061,可以采用液晶显示器、有机发光二极管等形式来配置显示面板9061。用户输入单元907包括触控面板9071以及其他输入设备9072中的至少一种。触控面板9071,也称为触摸屏。触控面板9071可包括触摸检测装置和触摸控制器两个部分。其他输入设备9072可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆,在此不再赘述。
本申请实施例中,射频单元901接收来自网络侧设备的下行数据后,可以传输给处理器910进行处理;另外,射频单元901可以向网络侧设备发送上行数据。通常,射频单元901包括但不限于天线、放大器、收发信机、耦合器、低噪声放大器、双工器等。
存储器909可用于存储软件程序或指令以及各种数据。存储器909可主要包括存储程序或指令的第一存储区和存储数据的第二存储区,其中,第一存储区可存储操作系统、至少一个功能所需的应用程序或指令(比如声音播放功能、图像播放功能等)等。此外,存储器909可以包括易失性存储器或非易失性存储器,或者,存储器909可以包括易失性和非易失性存储器两者。其中,非易失性存储器可以是只读存储器(Read-Only Memory,ROM)、可编程只读存储器(Programmable ROM,PROM)、可擦除可编程只读存储器(Erasable PROM,EPROM)、电可擦除可编程只读存储器(Electrically EPROM, EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory,RAM),静态随机存取存储器(Static RAM,SRAM)、动态随机存取存储器(Dynamic RAM,DRAM)、同步动态随机存取存储器(Synchronous DRAM,SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM,DDRSDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM,ESDRAM)、同步连接动态随机存取存储器(Synch link DRAM,SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM,DRRAM)。本申请实施例中的存储器909包括但不限于这些和任意其它适合类型的存储器。
处理器910可包括一个或多个处理单元;可选的,处理器910集成应用处理器和调制解调处理器,其中,应用处理器主要处理涉及操作系统、用户界面和应用程序等的操作,调制解调处理器主要处理无线通信信号,如基带处理器。可以理解的是,上述调制解调处理器也可以不集成到处理器910中。
其中,射频单元901,用于从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;
处理器910,用于应用所述配置信息配置所述时间序列预测模型。
本申请实施例还提供一种网络侧设备,包括处理器和通信接口,处理器用于实现上述预测信息的获取方法实施例的各个过程,通信接口用于与外部通信设备进行通信。该网络侧设备实施例与上述第二通信设备方法实施例对应,上述方法实施例的各个实施过程和实现方式均可适用于该网络侧设备实施例中,且能达到相同的技术效果。
具体地,本申请实施例还提供了一种网络侧设备。如图10所示,该网络侧设备1000包括:天线1001、射频装置1002、基带装置1003、处理器1004和存储器1005。天线1001与射频装置1002连接。在上行方向上,射频装置1002通过天线1001接收信息,将接收的信息发送给基带装置1003进行处理。在下行方向上,基带装置1003对要发送的信息进行处理,并发送给射频装置1002,射频装置1002对收到的信息进行处理后经过天线1001发送出去。
以上实施例中网络侧设备执行的方法可以在基带装置1003中实现,该基带装置1003包括基带处理器。
基带装置1003例如可以包括至少一个基带板,该基带板上设置有多个芯片,如图10所示,其中一个芯片例如为基带处理器,通过总线接口与存储器1005连接,以调用存储器1005中的程序,执行以上方法实施例中所示的网络设备操作。
该网络侧设备还可以包括网络接口1006,该接口例如为通用公共无线接口(common public radio interface,CPRI)。
具体地,本申请实施例的网络侧设备1000还包括:存储在存储器1005上并可在处理器1004上运行的指令或程序,处理器1004调用存储器1005中的指令或程序执行图7所示各模块执行的方法,并达到相同的技术效果,为避免重复,故不在此赘述。
本申请实施例还提供一种可读存储介质,所述可读存储介质上存储有程序或指令,该程序或指令被处理器执行时实现上述模型的构建方法实施例的各个过程,或实现上述预测信息的获取方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
其中,所述处理器为上述实施例中所述的终端中的处理器。所述可读存储介质,包括计算机可读存储介质,如计算机只读存储器ROM、随机存取存储器RAM、磁碟或者光盘等。
本申请实施例另提供了一种芯片,所述芯片包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现上述模型的构建方法实施例的各个过程,或实现上述预测信息的获取方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
应理解,本申请实施例提到的芯片还可以称为系统级芯片,系统芯片,芯片系统或片上系统芯片等。
本申请实施例另提供了一种计算机程序/程序产品,所述计算机程序/程序 产品被存储在存储介质中,所述计算机程序/程序产品被至少一个处理器执行以实现上述模型的构建方法实施例的各个过程,或实现上述预测信息的获取方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例还提供了一种模型的配置系统,包括:第一通信设备及第二通信设备,所述第一通信设备可用于执行如上所述的模型的构建方法的步骤,所述网络侧设备可用于执行如上所述的预测信息的获取方法的步骤。
需要说明的是,在本文中,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外,需要指出的是,本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能,还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能,例如,可以按不同于所描述的次序来执行所描述的方法,并且还可以添加、省去、或组合各种步骤。另外,参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以计算机软件产品的形式体现出来,该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端(可以是手机,计算机,服务器,空调器,或者网络设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述,但是本申请并不局限于上 述的具体实施方式,上述的具体实施方式仅仅是示意性的,而不是限制性的,本领域的普通技术人员在本申请的启示下,在不脱离本申请宗旨和权利要求所保护的范围情况下,还可做出很多形式,均属于本申请的保护之内。

Claims (39)

  1. 一种模型的构建方法,包括:
    第一通信设备从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;
    所述第一通信设备应用所述配置信息构建所述时间序列预测模型。
  2. 根据权利要求1所述的方法,其中,所述配置信息包括以下至少之一:
    核函数的第一配置信息,其中,所述第一配置信息包括:核函数的数量、核函数的超参数以及核函数的类型;
    第一任务标识,其中,所述第一任务标识用于指示所述第一配置信息相关联的任务;
    优化器选择信息;
    模型训练配置,其中,所述模型训练配置包括以下至少之一:第一训练窗的长度、模型训练的输入、模型训练的标签、第一训练窗样本采样的时间间隔;
    模型预测配置,其中,所述模型预测配置包括以下至少之一:第一预测窗的长度、预测的输入、预测的输出、第一预测窗样本采样的时间间隔;
    所述时间序列预测模型的生命周期;
    所述时间序列预测模型的计算模式,其中,所述计算模式包括:串行计算或并行计算。
  3. 根据权利要求2所述的方法,其中,所述优化器选择信息包括以下之一:
    梯度下降优化法的第一配置参数,其中,所述第一配置参数包括:初始值、步长、和终止条件;
    网格搜索优化法的第二配置参数,其中,所述第二配置参数包括:不同超参数网格搜索的上界、不同超参数网格搜索的下界、和不同超参数网络搜索的步长。
  4. 根据权利要求2所述的方法,其中,所述预测的输出包括:
    与所述预测输入的时间单位编号对应的标签信息;以及
    所述时间序列预测模型输出的预测方差和/或所述时间序列预测模型输出的预测误差。
  5. 根据权利要求2所述的方法,其中,所述时间序列预测模型的生命周期包括:生效时间、失效时间、以及运行时间。
  6. 根据权利要求1所述的方法,其中,在所述第一通信设备从第二通信设备获取时间序列预测模型的配置信息之后,所述方法还包括:
    所述第一通信设备向所述第二通信设备发送反馈信息,其中,所述反馈信息用于指示所述第一通信设备支持所述配置信息、或者所述反馈信息用于指示所述第一通信设备不支持所述配置信息。
  7. 根据权利要求6所述的方法,其中,所述反馈信息包括以下至少之一:
    第二指示信息,所述第二指示信息用于指示所述第一通信设备所支持的核函数,其中,所述第二指示信息指示的核函数为所述配置信息中配置的一个或多个核函数;
    第三指示信息,其中,所述第三指示信息用于指示所述第一通信设备具有任务预测的需求,或者,所述第三指示信息用于指示所述第一通信设备不具有所述任务预测的需求,所述第三指示信息指示的任务为所述配置信息配置的任务;
    第四指示信息,其中,所述第四指示信息用于指示所述第一通信设备支持目标优化器,或者,所述第四指示信息用于指示所述第一通信设备不支持目标优化器,所述目标优化器为所述配置信息配置的优化器;
    第五指示信息,其中,所述第五指示信息用于指示所述第一通信设备支持所述配置信息中的模型训练配置,或者,所述第五指示信息用于指示所述第一通信设备不支持所述配置信息中的模型训练配置;
    第六指示信息,其中,所述第六指示信息用于指示所述第一通信设备支 持所述配置信息中的模型预测配置,或者,所述第六指示信息用于指示所述第一通信设备不支持所述配置信息中的模型预测配置。
  8. 根据权利要求2所述的方法,其中,所述核函数为核函数列表中的至少一个核函数,其中,所述第一通信设备和所述第二通信设备中预先配置有所述核函数列表。
  9. 根据权利要求2所述的方法,其中,所述任务标识为任务列表中的至少一个任务,其中,所述第一通信设备和所述第二通信设备中预先配置有所述任务列表。
  10. 根据权利要求2所述的方法,其中,所述优化器选择信息中包括目标优化器标识,所述目标优化器标识为优化器列表中的至少一个优化器标识,所述优化器列表中包括优化器标识与参数配置的对应关系,所述第一通信设备和所述第二通信设备中预先配置有所述优化器列表。
  11. 根据权利要求2所述的方法,其中,所述模型训练配置还包括:
    对训练输入的第一处理方式,其中,所述第一处理方式包括:线性放缩处理或非线性放缩处理;
    对标签的第二处理方式,其中,所述第二处理方式包括:线性归一化处理或均值归一化处理。
  12. 根据权利要求2所述的方法,其中,所述计算模式配置由所述第一通信设备根据其计算能力和存储能力确定。
  13. 根据权利要求1所述的方法,其中,在第一通信设备从第二通信设备获取时间序列预测模型的配置信息之前,所述方法还包括:
    所述第一通信设备根据第二信息,向所述第二通信设备推荐所述时间序列预测模型的推荐配置。
  14. 根据权利要求13所述的方法,其中,所述推荐配置包括以下至少之一:
    推荐的核函数;
    推荐的模型训练配置;
    推荐的模型预测配置。
  15. 根据权利要求13所述的方法,其中,所述第二信息包括以下至少之一:
    待预测的第一信息的统计信息;
    所述第一信息的估计误差;
    所述估计误差的统计信息;
    模型预测误差;
    所述模型预测误差的统计信息;
    所述第一通信设备的移动性信息;
    噪声的统计信息;
    性能需求信息,其中,所述性能需求信息包括以下至少一项:预测精度的需求信息、处理时延的需求信息、计算时延的需求信息。
  16. 根据权利要求1至15任一项所述的方法,其中,在所述第一通信设备应用所述配置信息配置所述时间序列预测模型之后,所述方法还包括:
    所述第一通信设备向所述第二通信设备上报目标预测信息,其中,所述目标预测信息为使用所述时间序列预测模型对第一信息进行预测得到的预测信息,所述第一信息为时间序列相关的信息。
  17. 根据权利要求16所述的方法,其中,所述目标预测信息包括以下至少之一:
    预测得到的第一信息;
    预测所述第一信息对应的预测误差;
    第二任务标识;
    第二训练窗的配置相关信息;
    第二预测窗的配置相关信息。
  18. 根据权利要求17所述的方法,其中,所述第二训练窗的配置相关信 息包括以下至少之一:
    第二训练窗的长度;
    模型训练的输入;
    第一时间戳信息,其中,所述第一时间戳信息包括以下至少之一:第二训练窗的编号、第二训练窗的开始时间、第二训练窗的结束时间、第二训练窗样本采样的时间间隔。
  19. 根据权利要求17所述的方法,其中,所述第二预测窗的配置相关信息包括以下至少之一:
    第二预测窗的长度;
    模型预测的输入;
    第二时间戳信息,其中,所述第二时间戳信息包括以下至少之一:第二预测窗的编号、第二预测窗的开始时间、第二预测窗的结束时间、第二预测窗样本采样的时间间隔。
  20. 根据权利要求1至15任一项所述的方法,其中,在第一通信设备从第二通信设备获取时间序列预测模型的配置信息之前,所述方法还包括:
    所述第一通信设备向所述第二通信设备发送模型配置请求信息。
  21. 根据权利要求1所述的方法,其中,所述配置信息包括以下至少之一:
    所述时间序列预测模型的结构信息;
    所述时间序列预测模型的权值;
    所述时间序列预测模型的配置,其中,所述配置包括以下至少之一:优化器、损失函数;
    所述时间序列预测模型的优化器的状态信息。
  22. 根据权利要求1至15任一项所述的方法,其中,第一通信设备从第二通信设备获取时间序列预测模型的配置信息,包括以下之一:
    所述第一通信设备接收所述第二通信设备非周期性配置的所述配置信息;
    所述第一通信设备接收所述第二通信设备周期性配置的所述配置信息;
    所述第一通信设备与所述第二通信设备约定所述配置信息;
    所述第一通信设备确定所述配置信息,并向所述第二通信设备上报所述配置信息,接收到所述第二通信设备的确认信息后,获取所述配置信息。
  23. 一种预测信息的获取方法,包括:
    第二通信设备为第一通信设备配置时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;
    所述第二通信设备接收所述第一通信设备上报的目标预测信息,其中,所述目标预测信息为所述第一通信设备使用所述时间序列预测模型对第一信息进行预测得到的预测信息。
  24. 根据权利要求23所述的方法,其中,所述配置信息包括以下至少之一:
    核函数的第一配置信息,其中,所述第一配置信息包括:核函数的数量、核函数的超参数以及核函数的类型;
    任务标识,其中,所述任务标识用于指示所述第一配置信息相关联的任务;
    优化器选择信息;
    模型训练配置,其中,所述模型训练配置包括以下至少之一:第一训练窗的长度、模型训练的输入、模型训练的标签、第一训练窗样本采样的时间间隔;
    模型预测配置,其中,所述模型预测配置包括以下至少之一:预测窗的长度、预测的输入、预测的输出、预测窗样本采样的时间间隔;
    所述时间序列预测模型的生命周期;
    所述时间序列预测模型的计算模式配置,其中,所述计算模式配置包括:串行计算或并行计算。
  25. 根据权利要求23所述的方法,其中,在第二通信设备为第一通信设 备配置时间序列预测模型的配置信息之后,所述方法还包括:
    所述第二通信设备接收所述第一通信设备发送的反馈信息,其中,所述反馈信息用于指示所述第一通信设备支持所述配置信息、或者所述反馈信息用于指示所述第一通信设备不支持所述配置信息。
  26. 根据权利要求25所述的方法,其中,所述反馈信息包括以下至少之一:
    第二指示信息,所述第二指示信息用于指示所述第一通信设备所支持的核函数,其中,所述第二指示信息指示的核函数为所述配置信息中配置的一个或多个核函数;
    第三指示信息,其中,所述第三指示信息用于指示所述第一通信设备具有任务预测的需求,或者,所述第三指示信息用于指示所述第一通信设备不具有所述任务预测的需求,所述第三指示信息指示的任务为所述配置信息配置的任务;
    第四指示信息,其中,所述第四指示信息用于指示所述第一通信设备支持目标优化器,或者,所述第四指示信息用于指示所述第一通信设备不支持目标优化器,所述目标优化器为所述配置信息配置的优化器;
    第五指示信息,其中,所述第五指示信息用于指示所述第一通信设备支持所述配置信息中的模型训练配置,或者,所述第五指示信息用于指示所述第一通信设备不支持所述配置信息中的模型训练配置;
    第六指示信息,其中,所述第六指示信息用于指示所述第一通信设备支持所述配置信息中的模型预测配置,或者,所述第六指示信息用于指示所述第一通信设备不支持所述配置信息中的模型预测配置。
  27. 根据权利要求23所述的方法,其中,在第二通信设备为第一通信设备配置时间序列预测模型的配置信息之前,所述方法还包括:
    所述第二通信设备接收所述第一通信设备发送的所述时间序列预测模型的推荐配置。
  28. 根据权利要求27所述的方法,其中,所述推荐配置包括以下至少之一:
    推荐的核函数;
    推荐的模型训练配置;
    推荐的模型预测配置。
  29. 根据权利要求23所述的方法,其中,第二通信设备为第一通信设备配置时间序列预测模型的配置信息,包括:
    所述第二通信设备接收所述第一通信设备上报的第二信息;
    基于所述第二信息,为所述第一通信设备配置时间序列预测模型的配置信息;
    其中,所述第二信息包括以下至少之一:
    第一信息的统计信息;
    第一信息的估计误差;
    所述估计误差的统计信息;
    模型预测误差;
    所述模型预测误差的统计信息;
    所述第一通信设备的移动性信息;
    噪声的统计信息;
    性能需求信息,其中,所述性能需求信息包括以下至少一项:预测精度的需求信息、处理时延的需求信息、计算时延的需求信息。
  30. 根据权利要求23至29中任一项所述的方法,其中,所述目标预测信息包括以下至少之一:
    预测得到的第一信息;
    预测所述第一信息对应的预测误差;
    任务标识;
    第二训练窗配置的相关信息;
    第二预测窗的相关信息。
  31. 根据权利要求30所述的方法,其中,所述第二训练窗配置的相关信息包括以下至少之一:
    第二训练窗的长度;
    模型训练的输入;
    第一时间戳信息,其中,所述第一时间戳信息包括以下至少之一:第二训练窗的编号、第二训练窗的开始时间、第二训练窗的结束时间、第二训练窗样本采样的时间间隔。
  32. 根据权利要求30所述的方法,其中,所述第二预测窗的相关信息包括以下至少之一:
    第二预测窗的长度;
    模型预测的输入;
    第二时间戳信息,其中,所述第二时间戳信息包括以下至少之一:第二预测窗的编号、第二预测窗的开始时间、第二预测窗的结束时间、第二预测窗样本采样的时间间隔。
  33. 根据权利要求23至32任一项所述的方法,其中,在第二通信设备为第一通信设备配置时间序列预测模型的配置信息之前,所述方法还包括:
    接收所述第一通信设备发送的模型配置请求信息。
  34. 根据权利要求23所述的方法,其中,所述配置信息包括以下至少之一:
    所述时间序列预测模型的结构信息;
    所述时间序列预测模型的权值;
    所述时间序列预测模型的配置,其中,所述配置包括以下至少之一:优化器、损失函数;
    所述时间序列预测模型的优化器的状态信息。
  35. 根据权利要求23至34任一项所述的方法,其中,所述第二通信设 备为第一通信设备配置时间序列预测模型的配置信息,包括以下之一:
    所述第二通信设备非周期性为第一通信设备配置时间序列预测模型的配置信息;
    所述第二通信设备周期性为第一通信设备配置时间序列预测模型的配置信息;
    所述第二通信设备与所述第一通信设备约定所述配置信息;
    所述第二通信设备所述第二通信设备上报所述配置信息,向所述第一通信设备发送确认信息。
  36. 一种模型的构建装置,包括:
    获取模块,用于从第二通信设备获取时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测时间序列相关的信息;
    应用模块,用于应用所述配置信息构建所述时间序列预测模型。
  37. 一种预测信息的获取装置,包括:
    配置模块,用于为第一通信设备配置时间序列预测模型的配置信息,其中,所述时间序列预测模型用于预测第一信息,所述第一信息为时间序列相关的信息;
    接收模块,用于接收所述第一通信设备上报的目标预测信息,其中,所述目标预测信息为所述第一通信设备使用所述时间序列预测模型对所述第一信息进行预测得到的预测信息。
  38. 一种通信设备,包括处理器和存储器,所述存储器存储可在所述处理器上运行的程序或指令,所述程序或指令被所述处理器执行时实现如权利要求1至35任一项所述的方法的步骤。
  39. 一种可读存储介质,所述可读存储介质上存储程序或指令,所述程序或指令被处理器执行时实现如权利要求1至35任一项所述的方法的步骤。
PCT/CN2022/143672 2021-12-30 2022-12-29 模型的构建方法、装置及通信设备 WO2023125880A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/746,032 US20240346113A1 (en) 2021-12-30 2024-06-17 Model construction method and apparatus, and communication device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111662864.6 2021-12-30
CN202111662864.6A CN116415476A (zh) 2021-12-30 2021-12-30 模型的构建方法、装置及通信设备

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/746,032 Continuation US20240346113A1 (en) 2021-12-30 2024-06-17 Model construction method and apparatus, and communication device

Publications (1)

Publication Number Publication Date
WO2023125880A1 true WO2023125880A1 (zh) 2023-07-06

Family

ID=86998105

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/143672 WO2023125880A1 (zh) 2021-12-30 2022-12-29 模型的构建方法、装置及通信设备

Country Status (3)

Country Link
US (1) US20240346113A1 (zh)
CN (1) CN116415476A (zh)
WO (1) WO2023125880A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117272234A (zh) * 2023-11-21 2023-12-22 南京邮电大学 一种基于稠密-稀疏的时空transformer模型的WiFi行为识别方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160034823A1 (en) * 2014-07-31 2016-02-04 Collision Communications, Inc. Methods, Systems, And Computer Program Products For Optimizing A Predictive Model For Mobile Network Communications Based On Historical Context Information
CN105744625A (zh) * 2014-12-12 2016-07-06 索尼公司 用于无线通信的装置和方法
WO2018202297A1 (en) * 2017-05-04 2018-11-08 Huawei Technologies Co., Ltd. Network computing device, method and system for time-series forecasting based on a moving-median algorithm
CN110543973A (zh) * 2019-08-06 2019-12-06 阿里巴巴集团控股有限公司 数据预测、数据预测结果的调整方法、装置及设备
CN113268403A (zh) * 2021-05-25 2021-08-17 中国联合网络通信集团有限公司 时间序列的分析预测方法、装置、设备及存储介质
CN113660676A (zh) * 2021-07-20 2021-11-16 山东师范大学 基站流量预测方法、系统、存储介质及设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160034823A1 (en) * 2014-07-31 2016-02-04 Collision Communications, Inc. Methods, Systems, And Computer Program Products For Optimizing A Predictive Model For Mobile Network Communications Based On Historical Context Information
CN105744625A (zh) * 2014-12-12 2016-07-06 索尼公司 用于无线通信的装置和方法
WO2018202297A1 (en) * 2017-05-04 2018-11-08 Huawei Technologies Co., Ltd. Network computing device, method and system for time-series forecasting based on a moving-median algorithm
CN110543973A (zh) * 2019-08-06 2019-12-06 阿里巴巴集团控股有限公司 数据预测、数据预测结果的调整方法、装置及设备
CN113268403A (zh) * 2021-05-25 2021-08-17 中国联合网络通信集团有限公司 时间序列的分析预测方法、装置、设备及存储介质
CN113660676A (zh) * 2021-07-20 2021-11-16 山东师范大学 基站流量预测方法、系统、存储介质及设备

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117272234A (zh) * 2023-11-21 2023-12-22 南京邮电大学 一种基于稠密-稀疏的时空transformer模型的WiFi行为识别方法
CN117272234B (zh) * 2023-11-21 2024-01-30 南京邮电大学 一种基于稠密-稀疏的时空transformer模型的WiFi行为识别方法

Also Published As

Publication number Publication date
US20240346113A1 (en) 2024-10-17
CN116415476A (zh) 2023-07-11

Similar Documents

Publication Publication Date Title
CN109918684B (zh) 模型训练方法、翻译方法、相关装置、设备及存储介质
US11169827B2 (en) Resource loading at application startup using attributes of historical data groups
US10070246B2 (en) Electronic device and method for transmitting information
CN110069715B (zh) 一种信息推荐模型训练的方法、信息推荐的方法及装置
US20240346113A1 (en) Model construction method and apparatus, and communication device
EP3131317A1 (en) Method and apparatus for providing location information
WO2024148870A1 (zh) 一种调频方法、装置、电子设备及可读存储介质
Chakrabarti Deep learning based offloading for mobile augmented reality application in 6G
Nabhan et al. Correlation-based dynamic sampling for online high dimensional process monitoring
US20160134743A1 (en) Method and apparatus for managing function of wearable device
US10045278B2 (en) Electronic device and method for searching for access point in local area network at electronic device
CN111090877A (zh) 数据生成、获取方法及对应的装置、存储介质
CN109583583B (zh) 神经网络训练方法、装置、计算机设备及可读介质
CN113704008A (zh) 一种异常检测方法、问题诊断方法和相关产品
WO2023125879A1 (zh) 数据处理的方法、装置及通信设备
CN115866726A (zh) 用于使网络通信量对齐以改善功耗的方法和装置
US10552741B1 (en) Computing smoothed posterior distribution of confidential data
WO2023213270A1 (zh) 模型训练处理方法、装置、终端及网络侧设备
WO2023088387A1 (zh) 信道预测方法、装置、ue及系统
WO2023125855A1 (zh) 模型更新方法及通信设备
WO2023174253A1 (zh) Ai模型的处理方法及设备
WO2024169794A1 (zh) 模型监督方法及通信设备
WO2023143415A1 (zh) 信息反馈方法、接收方法、装置、终端及网络侧设备
WO2024140445A1 (zh) 模型信息传输方法、装置及设备
WO2024027576A1 (zh) 一种ai网络模型的性能监督方法、装置和通信设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22915129

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2022915129

Country of ref document: EP

Effective date: 20240730