CN115830709A - Action recognition method based on federal edge learning, server and electronic equipment - Google Patents
Action recognition method based on federal edge learning, server and electronic equipment Download PDFInfo
- Publication number
- CN115830709A CN115830709A CN202211474539.1A CN202211474539A CN115830709A CN 115830709 A CN115830709 A CN 115830709A CN 202211474539 A CN202211474539 A CN 202211474539A CN 115830709 A CN115830709 A CN 115830709A
- Authority
- CN
- China
- Prior art keywords
- information
- edge
- model
- model parameters
- equipment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses an action recognition method based on federal edge learning, a server and electronic equipment, which are applied to an edge server, wherein the method comprises the following steps: receiving a training request and equipment information sent by at least one edge equipment; determining a resource allocation strategy for allocating the edge device according to the device information; sending a resource allocation strategy and initial model parameters obtained from a pre-configured artificial intelligence model to the edge device according to the training request so as to acquire information and a plurality of global model parameters; training the artificial intelligent model according to the plurality of global model parameters and the acquired information to obtain a target recognition model; and sending the target identification model to the edge equipment so that the edge equipment inputs the collected sample to be detected into the target identification model for action identification and outputs the target action information of the sample to be detected. According to the embodiment of the invention, the federal learning training efficiency can be improved and the accuracy of action recognition can be improved under the condition of resource constraint.
Description
Technical Field
The invention belongs to the technical field of computers, and particularly relates to an action recognition method based on federal edge learning, a server, terminal equipment, electronic equipment and a storage medium.
Background
With the explosion of information and communication technologies, more and more intelligent applications consider using artificial intelligence technology, and the ever-increasing edge devices such as various mobile devices and sensors generate a large amount of data at the edge of the network. Thus, a new learning paradigm, namely edge learning, is brought forward. In edge learning, the federate edge learning machine learning framework is concerned with and widely used due to the advantages of fully utilizing local computing resources and protecting local data privacy, so the federate edge learning can be applied to various fields such as image recognition, image classification and action recognition.
For the field of action recognition, edge devices in federal edge learning need to sense action information of human bodies in the environment to obtain data for training, model updating is calculated by using local computing power, finally, the local model updating is uploaded to an edge server through an uplink channel, and the edge server broadcasts a global model to each edge device through a downlink channel to complete the process of identifying the action information in the environment.
However, in the conventional federal edge learning-based action recognition method, when resources such as power and time of edge devices in an edge network are limited, the federal edge learning training efficiency is slow, and the resource utilization rate is low, so that the action recognition efficiency is reduced.
Disclosure of Invention
The invention aims to at least solve one of the technical problems in the prior art, and provides an action recognition method based on federal edge learning, a server, a terminal device, an electronic device and a storage medium, which can improve the federal learning training efficiency and improve the accuracy of action recognition under the condition of resource constraint.
In a first aspect, the present invention provides an action recognition method based on federal edge learning, which is applied to an edge server, where the edge server is in communication connection with a plurality of edge devices, and the method includes:
receiving a training request and equipment information sent by at least one piece of edge equipment, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body;
determining a resource allocation strategy for allocating the edge device according to the device information;
sending the resource allocation strategy and initial model parameters obtained from a pre-configured artificial intelligence model to the edge devices according to the training request, so that each edge device obtains acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters, wherein the acquisition information is obtained by the edge devices according to the resource allocation strategy through acquiring action information of a human body, and the model parameters are obtained by the edge devices according to the initial model parameters through updating;
training the artificial intelligence model according to the plurality of model parameters and the acquired information to obtain a target recognition model;
and sending the target identification model to the edge equipment so that the edge equipment inputs the collected sample to be detected into the target identification model for action identification and outputs target action information of the sample to be detected.
According to the action recognition method based on the federal edge learning provided by the embodiment of the invention, at least the following beneficial effects are achieved: the method comprises the steps of firstly, receiving a training request and equipment information sent by at least one edge device to obtain the acquisition capacity of the edge device to the action information of a human body, then determining a resource allocation strategy for allocating the edge device according to the equipment information to enable the edge device to converge under a specific strategy, then sending the resource allocation strategy and initial model parameters obtained from a pre-configured artificial intelligence model to the edge device according to the training request to enable each edge device to calculate according to the resource allocation strategy and the initial model parameters to obtain acquired information and a plurality of global model parameters, thereby improving the resource utilization efficiency of the edge device, then training the artificial intelligence model according to the plurality of global model parameters and the acquired information to obtain a target identification model, and finally sending the target identification model to the edge device to enable the edge device to input the acquired sample to be detected into the target identification model for action identification, and output the target action information of the sample to be detected, thereby realizing the identification accuracy of the action information of the human body, and reasonably allocating resources between perception, communication and calculation under the given resource allocation strategy.
According to some embodiments of the invention, the device information comprises a pilot signal, device parameter information, and device power information; the determining a resource allocation policy for allocating the edge device according to the device information includes:
determining channel state information according to the pilot signal, wherein the channel state information comprises a fading factor;
calculating the fading factor, the equipment parameter information and the equipment power information based on a preset grid search algorithm to obtain the communication transmitting power of the edge equipment;
inputting the communication transmitting power into the edge equipment for calculation to obtain a perception data parameter;
inputting the perception data parameters into a preset training function for calculation to obtain the quantity of perception samples;
and generating the resource allocation strategy for allocating the edge device according to the sensing sample number and the communication transmission power.
According to some embodiments of the invention, the device power information comprises a transmit power and a transmit power; the calculating the fading factor, the device parameter information, and the device power information based on a preset grid search algorithm to obtain the communication transmission power of the edge device includes:
calculating a preset exponential integral function, communication bandwidth, noise power spectral density information, the fading factor and the transmission power to obtain a channel capacity function;
and calculating the equipment parameter information, the transmitting power and the channel capacity function according to the grid search algorithm to obtain the communication transmitting power of the edge equipment.
According to some embodiments of the present invention, the training the artificial intelligence model according to the plurality of model parameters and the collected information to obtain a target recognition model includes:
carrying out average operation on the plurality of model parameters to obtain global model parameters;
training the artificial intelligence model according to the acquired information to obtain a loss function of the artificial intelligence model;
comparing the loss function of the artificial intelligence model with a preset threshold constraint condition to obtain a comparison result;
when the comparison result is that the function value of the loss function does not meet the threshold constraint condition, sending the global model parameter to the edge device so that the edge device performs multi-step local updating based on the global model parameter until the updated function value of the loss function meets the threshold constraint condition, and obtaining the target identification model according to the updated artificial intelligence model;
and when the comparison result is that the function value of the loss function meets the threshold constraint condition, obtaining the target identification model according to the trained artificial intelligence model.
In a second aspect, the present invention provides an action recognition method based on federal edge learning, which is applied to an edge device, wherein an edge server is in communication connection with a plurality of edge devices, and the method includes:
sending a training request and equipment information to the edge server to enable the edge server to generate a resource allocation strategy and initial model parameters according to the training request and the equipment information, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body in the environment, and the initial model parameters are used for representing the model performance of an artificial intelligent model in the edge server;
acquiring acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters, wherein the acquisition information is used for representing the acquisition capacity of the acquisition information on the action information of the human body under the resource allocation strategy, and the model parameters are used for representing the updating capacity of the initial model parameters under the initial model parameters;
sending the acquisition information and the plurality of model parameters to the edge server so that the edge server trains the artificial intelligence model according to the acquisition information and the plurality of model parameters to obtain a target recognition model;
collecting motion information of a human body to obtain a sample to be detected;
and receiving the target identification model sent by the edge server, and performing action identification on the sample to be detected according to the target identification model to obtain target action information.
According to some embodiments of the invention, the resource allocation policy comprises a perceived sample number; the obtaining of the acquisition information and the plurality of model parameters according to the resource allocation strategy and the initial model parameters comprises:
acquiring motion information of a human body according to the number of the perception samples to obtain the acquired information;
and performing multi-step local updating based on the initial model parameters to obtain a plurality of model parameters.
According to some embodiments of the invention, further comprising:
under the condition that the function value of the loss function of the artificial intelligence model does not meet the threshold constraint condition, receiving the global model parameter sent by the edge server, and performing multi-step local updating based on the global model parameter to obtain an updating result;
and sending the updating result to the edge server until the edge server determines that the function value of the loss function meets the threshold constraint condition according to the updating result.
In a third aspect, the present invention provides a server, comprising:
the information receiving module is used for receiving a training request and equipment information sent by at least one piece of edge equipment, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body;
a policy determining module, configured to determine, according to the device information, a resource allocation policy for allocating the edge device;
a policy sending module, configured to send the resource allocation policy and an initial model parameter obtained from a preconfigured artificial intelligence model to the edge device according to the training request, so that each edge device obtains acquisition information and multiple model parameters according to the resource allocation policy and the initial model parameter, where the acquisition information is obtained by acquiring, by the edge device, action information of a human body according to the resource allocation policy, and the model parameters are obtained by updating, by the edge device, according to the initial model parameter;
the model training module is used for training the artificial intelligence model according to the plurality of model parameters and the acquisition information to obtain a target recognition model;
and the model sending module is used for sending the target identification model to the edge equipment so that the edge equipment inputs the collected sample to be detected into the target identification model for action identification and outputs the target action information of the sample to be detected.
In a fourth aspect, the present invention provides a terminal device, including:
the information sending module is used for sending a training request and equipment information to the edge server so that the edge server can generate a resource allocation strategy and initial model parameters according to the training request and the equipment information, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body in the environment, and the initial model parameters are used for representing the model performance of an artificial intelligent model in the edge server;
the parameter calculation module is used for obtaining acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters, wherein the acquisition information is used for representing the acquisition capacity of the acquisition information of the human body under the resource allocation strategy, and the model parameters are used for representing the updating capacity of the initial model parameters under the initial model parameters;
the parameter sending module is used for sending the acquisition information and the plurality of model parameters to the edge server so that the edge server trains the artificial intelligence model according to the acquisition information and the plurality of model parameters to obtain a target recognition model;
the information acquisition module is used for acquiring the action information of the human body to obtain a sample to be detected;
and the target determining module is used for receiving the target identification model sent by the edge server and carrying out action identification on the sample to be detected according to the target identification model to obtain target action information.
In a fifth aspect, the present invention provides an electronic device, where the electronic device includes a memory and a processor, where the memory stores a computer program, and the processor implements the method for identifying an action based on federated edge learning according to the first aspect and/or the second aspect when executing the computer program.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the example serve to explain the principles of the invention and not to limit the invention.
Fig. 1 is a flowchart of an action recognition method based on federal edge learning according to an embodiment of the present invention;
FIG. 2 is a flowchart of a specific method of step S102 in FIG. 1;
FIG. 3 is a flowchart of a specific method of step S202 in FIG. 2;
FIG. 4 is a flowchart illustrating a detailed method of step S104 in FIG. 1;
FIG. 5 is a flowchart of a method for federated edge learning based action recognition according to another embodiment of the present invention;
FIG. 6 is a flowchart of a detailed method of step S502 in FIG. 5;
FIG. 7 is a flowchart of a method for identifying actions based on federated edge learning according to another embodiment of the present invention;
FIG. 8 is a block diagram of a server according to an embodiment of the present invention;
fig. 9 is a schematic structural diagram of a terminal device according to an embodiment of the present invention;
fig. 10 is a schematic hardware structure diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
It should be noted that, although a logical order is illustrated in the flowcharts, in some cases, the steps illustrated or described may be performed in an order different from that in the flowcharts. The terms first, second and the like in the description and in the claims, and the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
The invention provides an action recognition method based on federal marginal learning, a server, terminal equipment, electronic equipment and a storage medium, wherein the joint participation of three processes of communication, perception and calculation is involved in the process of a federal marginal learning training model. In each round of iterative training, the edge device needs to exchange information with the edge server through wireless communication, and the edge device has functions of communication, calculation and environment perception at the same time.
The embodiments of the present invention will be further explained with reference to the drawings.
Referring to fig. 1, fig. 1 is a flowchart of an action recognition method based on federal edge learning according to an embodiment of the present invention, which is applied to an edge server, where the edge server is communicatively connected to a plurality of edge devices, and the action recognition method includes, but is not limited to, steps S101 to S105.
Step S101: receiving a training request and equipment information sent by at least one edge equipment;
it should be noted that the device information is used to characterize the acquisition capability of the edge device on the motion information of the human body.
In some embodiments, the edge server receives a training request and device information sent by at least one edge device, so as to respond to the edge device according to the training request, thereby realizing real-time communication between the edge server and the edge device.
Step S102: determining a resource allocation strategy for allocating the edge device according to the device information;
in some embodiments, a resource allocation policy for allocating the edge device is determined according to the device information, so that resource limitation can be performed on information such as communication transmission power of the edge device and the number of samples that need to be sensed in each round, and resource utilization rate of federal edge learning is improved under the condition that resources are limited.
Step S103: sending a resource allocation strategy and initial model parameters obtained from a pre-configured artificial intelligence model to the edge equipment according to the training request, so that each edge equipment obtains acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters;
it should be noted that the acquired information is obtained by the edge device acquiring the motion information of the human body according to the resource allocation policy, and the model parameters are obtained by the edge device updating according to the initial model parameters.
In some embodiments, the method includes the steps of determining that an edge server allows an edge device to join a trained edge device according to a training request, then sending a resource allocation strategy and initial model parameters obtained from a pre-configured artificial intelligence model to the edge device applying for passing, so that each edge device acquires action information of a human body according to the resource allocation strategy to obtain acquired information, and training and updating the initial model parameters to obtain model parameters, thereby facilitating subsequent training of the artificial intelligence model, realizing training of the artificial intelligence model under the condition that resources are limited, and realizing full utilization of resources.
It should be noted that the initial model parameters include, but are not limited to, parameters including weight values, bias values, and the like of the neural network model, and the embodiment is not particularly limited.
Step S104: training the artificial intelligent model according to the plurality of model parameters and the acquired information to obtain a target recognition model;
in some embodiments, the artificial intelligence model is trained according to a plurality of model parameters and the collected information, so that the collection capability and the recognition capability of the artificial intelligence model on the action information are improved, and the trained target recognition model is obtained.
Step S105: and sending the target identification model to the edge equipment so that the edge equipment inputs the collected sample to be detected into the target identification model for action identification and outputs the target action information of the sample to be detected.
In some embodiments, the target recognition model is sent to the edge device, so that the edge device can perform motion recognition on the collected sample to be detected according to the target recognition model, and output target motion information of the sample to be detected, thereby completing accurate recognition of human motion and improving recognition accuracy of the motion information of the human body.
In steps S101 to S105 provided in this embodiment, first, a training request and device information sent by at least one edge device are received, so as to obtain an acquisition capability of the edge device for motion information of a human body, then a resource allocation policy for allocating the edge device is determined according to the device information, so that the edge device can converge under a specific policy, then, the resource allocation policy and initial model parameters obtained from a preconfigured artificial intelligence model are sent to the edge device according to the training request, so that each edge device can perform calculation according to the resource allocation policy and the initial model parameters, so as to obtain acquisition information and a plurality of global model parameters, so as to improve resource utilization efficiency of the edge device, then, the artificial intelligence model is trained according to the plurality of global model parameters and the acquisition information, so as to obtain a target identification model, and finally, the target identification model is sent to the edge device, so that the edge device inputs the acquired sample to be detected into the target identification model for motion identification, and outputs target motion information of the sample to be detected, thereby achieving accuracy of recognition of the motion information of the human body, and achieving reasonable resource allocation between the sensing, communication policy, and the calculation under the given resource allocation.
Referring to fig. 2, fig. 2 is a flowchart of a specific method of step S102 in fig. 1, and step S102 includes, but is not limited to, steps S201 to S205.
It should be noted that the device information includes a pilot signal, device parameter information, and device power information.
Step S201: determining channel state information according to the pilot signal;
it should be noted that the channel state information includes a fading factor phi k 。
In some embodiments, the channel state information of the communication channel between different edge devices and the edge server is determined according to the pilot information, so that the attenuation amount of the signal in the communication channel can be obtained, and the calculation accuracy is improved.
Step S202: calculating a fading factor, equipment parameter information and equipment power information based on a preset grid search algorithm to obtain the communication transmitting power of the edge equipment;
in some embodiments, the mesh is based on presetsThe lattice search algorithm calculates the fading factor, the equipment parameter information and the equipment power information to obtain the communication transmitting power of the edge equipment
The communication transmitting power is the optimal communication transmitting power, so that the optimal communication transmitting power in the communication process of the edge device and the edge server can be obtained, and the subsequent edge device can be updated locally.
It should be noted that the device parameter information includes the device power E max And a Central Processing Unit (CPU) frequency f cpu Time T required for sensing a sample 0 The present embodiment is not particularly limited.
It can be understood that the grid search method is an exhaustive search method for specifying parameter values, and the parameters of the estimation function are optimized by a cross validation method to obtain an optimal learning algorithm.
Step S203: inputting the communication transmitting power into edge equipment for calculation to obtain a sensing data parameter;
in some embodiments, the communication transmission power is input to the edge device for calculation to obtain the perception data parameter
Wherein the sensing data parameter is the maximum total sensing data quantity of the edge device.
It should be noted that the formula for obtaining the sensing data parameter through specific calculation is shown as the following formula (1):
wherein, tau is the local update step number of the edge device, C k (p c，k ) Is the ergodic capacity of the communication channel between the edge device and the server, v is the unit effort consumed to perform the random gradient descent, theta is the effective switched capacitance coefficient,
for communication transmit power, R is the upper limit on the number of training rounds, D b For the size of the artificial intelligence model, T max Is the upper training time limit.
Step S204: inputting the perception data parameters into a preset training function for calculation to obtain the quantity of perception samples;
in some embodiments, the perceptual data parameters are
And inputting a preset training function for calculation to obtain the number of perception samples, so that the number of samples which need to be perceived by the edge equipment in each round can be obtained, the number of samples which need to be perceived by the edge equipment under a resource allocation strategy can be conveniently obtained subsequently, and the accuracy of the model training subsequently is improved.
It should be noted that, the specific process of inputting the perceptual data parameters into the training function for calculation is shown in the following formula (2):
wherein r is the number of rounds,
according to the method, the compound is obtained.
Step S205: and generating a resource allocation strategy for allocating the edge device according to the sensing sample number and the communication transmission power.
In some embodiments, a resource allocation policy for allocating the edge device is generated according to the sensing sample number and the communication transmission power, so that the optimal transmission power of the edge device and the number of samples to be sensed in each round can be limited through the resource allocation policy, and the training speed of federal edge learning can be improved subsequently under the condition that resources such as power, energy and time of the edge device in an edge network are limited.
Referring to fig. 3, fig. 3 is a flowchart of a specific method of step S202 in fig. 2, and step S202 includes, but is not limited to, steps S301 to S302.
It should be noted that the device power information includes the transmission power P s,k And transmission power
Wherein, the transmission power is the equipment sensing transmission power, and the transmission power is the highest communication transmission power.
Step S301: calculating a preset exponential integral function, communication bandwidth, noise power spectral density information, fading factors and transmission power to obtain a channel capacity function;
in some embodiments, the communication bandwidth B is calculated by applying a predetermined exponential integration function Ei (-) to 0 Noise power spectral density information N 0 The fading factor phi k And calculating the transmission power
Obtaining a channel capacity function, thereby obtaining the ergodic capacity of the communication channel between the edge device and the edge server, wherein the specific calculation is as shown in the following formula (3):
step S302: and calculating the equipment parameter information, the transmitting power and the channel capacity function according to a grid search algorithm to obtain the communication transmitting power of the edge equipment.
In some embodiments, the device parameter information, the transmission power, and the channel capacity function are calculated according to a grid search algorithm to obtain the communication transmission power of the edge device, so as to obtain the optimal communication transmission power of the edge device, where the specific calculation is as shown in the following formula (4):
referring to fig. 4, fig. 4 is a flowchart of a specific method of step S104 in fig. 1, and step S104 includes, but is not limited to, steps S401 to S405.
Step S401: carrying out average operation on the plurality of model parameters to obtain global model parameters;
in some embodiments, after receiving the model parameters sent by the edge device, averaging the received multiple model parameters to obtain global model parameters.
Step S402: training the artificial intelligence model according to the acquired information to obtain a loss function of the artificial intelligence model;
step S403: comparing the loss function of the artificial intelligence model with a preset threshold constraint condition to obtain a comparison result;
step S404: when the comparison result is that the function value of the loss function does not meet the threshold constraint condition, sending the global model parameter to the edge device so that the edge device can perform multi-step local updating based on the global model parameter until the updated function value of the loss function meets the threshold constraint condition, and obtaining a target identification model according to the updated artificial intelligence model;
step S405: and when the comparison result is that the function value of the loss function meets the threshold constraint condition, obtaining a target recognition model according to the trained artificial intelligence model.
In steps S401 to S405 of some embodiments, the edge server first aggregates a plurality of model parameters to obtain global model parameters, then trains the artificial intelligence model according to the acquired information to obtain a loss function of the artificial intelligence model, so as to subsequently determine a condition for finishing the training, then compares the loss function of the artificial intelligence model with a preset threshold constraint condition to obtain a comparison result, wherein when the comparison result is that the function value of the loss function does not satisfy the threshold constraint condition, the edge device continues to send the global model parameters to perform multi-step local update based on the global model parameters until the function value of the updated loss function satisfies the threshold constraint condition, and obtains a target identification model according to the updated artificial intelligence model; when the comparison result is that the function value of the loss function meets the threshold constraint condition, the target recognition model can be directly obtained according to the trained artificial intelligence model, and the training process of the artificial intelligence model is completed, so that the recognition capability of the artificial intelligence model on the action information is improved, and the accuracy of action information recognition is improved.
Referring to fig. 5, fig. 5 is a flowchart of an action recognition method based on federal edge learning according to another embodiment of the present invention, which is applied to an edge device, where an edge server is connected to a plurality of edge devices in a communication manner, and the action recognition method includes, but is not limited to, steps S501 to S505.
Step S501: sending a training request and equipment information to an edge server so that the edge server generates a resource allocation strategy and initial model parameters according to the training request and the equipment information;
it should be noted that the device information is used to represent the collection capability of the edge device on the motion information of the human body in the environment, and the initial model parameters are used to represent the model performance of the artificial intelligence model in the edge server.
In some embodiments, a training request and device information are sent to an edge server, so that the edge server can make corresponding feedback according to the training request, and can generate a resource allocation policy and initial model parameters corresponding to the edge device according to the device information.
It should be noted that the initial model parameters are initial model parameters obtained by the edge server from the preconfigured artificial intelligence model.
Step S502: acquiring acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters;
in some embodiments, the acquisition information and the plurality of model parameters are obtained according to the resource allocation strategy and the initial model parameters, so that the resources are fully utilized, and the convergence speed is accelerated under the condition of given resource constraint.
Step S503: sending the acquisition information and the plurality of model parameters to an edge server so that the edge server trains the artificial intelligent model according to the acquisition information and the plurality of model parameters to obtain a target recognition model;
in some embodiments, the collected information and the plurality of model parameters are sent to the edge server, so that the edge server trains the artificial intelligence model according to the collected information and the plurality of model parameters to obtain the target recognition model, thereby improving the recognition capability of the target recognition model on the action information of the human body.
Step S504: collecting motion information of a human body to obtain a sample to be detected;
step S505: and receiving a target identification model sent by the edge server, and performing action identification on the sample to be detected according to the target identification model to obtain target action information.
In some embodiments, after the edge server trains to obtain the target recognition model, the edge server receives the target recognition model sent by the edge server, collects motion information of a human body to obtain a sample to be detected, inputs the sample to be detected into the target recognition model to perform motion recognition, thereby obtaining target motion information, and realizes testing of the target recognition model.
Referring to fig. 6, fig. 6 is a flowchart of a specific method of step S502 in fig. 5, and step S502 includes, but is not limited to, steps S601-S602.
It should be noted that the resource allocation policy includes a sensing sample number, where the sensing sample number is a number of samples that the edge device needs to sense in this round.
Step S601: acquiring motion information of a human body according to the number of the perception samples to obtain acquired information;
in some embodiments, motion information of a human body needing to be sensed is acquired according to the number of sensing samples, and acquired information is obtained.
Step S602: and performing multi-step local updating based on the initial model parameters to obtain a plurality of model parameters.
In some embodiments, the plurality of model parameters are obtained by performing a plurality of steps of local update based on the initial model parameters, wherein the local update based on the initial model parameters is performed by a small batch of stochastic gradient descent update, which facilitates subsequent training of the artificial intelligence model.
Referring to fig. 7, fig. 7 is a flowchart of an action recognition method based on federated edge learning according to another embodiment of the present invention, which includes but is not limited to steps S701 to S702.
Step S701: under the condition that the function value of the loss function of the artificial intelligence model does not meet the threshold constraint condition, receiving the global model parameter sent by the edge server, and performing multi-step local updating based on the global model parameter to obtain an updating result;
step S702: and sending the updating result to the edge server until the edge server determines that the function value of the loss function meets the threshold constraint condition according to the updating result.
In some embodiments, under the condition that the function value of the loss function of the artificial intelligent model does not meet the threshold constraint condition, receiving the global model parameter sent by the edge server, performing multi-step local updating again based on the global model parameter to obtain an updating result, and sending the updating result to the edge server until the edge server determines that the function value of the loss function meets the threshold constraint condition according to the updating result, thereby completing the training process of the artificial intelligent model.
Referring to fig. 8, fig. 8 is a schematic structural diagram of a server according to an embodiment of the present invention, which may implement the method for identifying an action based on federal edge learning, where the server includes:
an information receiving module 801, configured to receive a training request and device information sent by at least one edge device, where the device information is used to represent an acquisition capability of the edge device on motion information of a human body;
a policy determining module 802, configured to determine, according to the device information, a resource allocation policy for allocating the edge device;
a policy sending module 803, configured to send a resource allocation policy and an initial model parameter obtained from a preconfigured artificial intelligence model to edge devices according to a training request, so that each edge device obtains acquisition information and a plurality of model parameters according to the resource allocation policy and the initial model parameter, where the acquisition information is obtained by the edge devices acquiring motion information of a human body according to the resource allocation policy, and the model parameters are obtained by the edge devices updating according to the initial model parameter;
the model training module 804 is used for training the artificial intelligent model according to a plurality of model parameters and the acquired information to obtain a target recognition model;
the model sending module 805 is configured to send the target identification model to the edge device, so that the edge device inputs the collected sample to be detected into the target identification model to perform motion identification, and outputs target motion information of the sample to be detected.
Referring to fig. 9, fig. 9 is a schematic structural diagram of a terminal device according to an embodiment of the present invention, which can implement the above action recognition method based on federal edge learning, where the terminal device includes:
an information sending module 901, configured to send a training request and device information to an edge server, so that the edge server generates a resource allocation policy and an initial model parameter according to the training request and the device information, where the device information is used to characterize an acquisition capability of the edge device on motion information of a human body in an environment, and the initial model parameter is used to characterize a model performance of an artificial intelligence model in the edge server;
a parameter calculation module 902, configured to obtain acquisition information and a plurality of model parameters according to the resource allocation policy and the initial model parameters, where the acquisition information is used to characterize acquisition capability of the human body for action information under the resource allocation policy, and the model parameters are used to characterize update capability of the initial model parameters under the initial model parameters;
a parameter sending module 903, configured to send the acquisition information and the plurality of model parameters to an edge server, so that the edge server trains the artificial intelligence model according to the acquisition information and the plurality of model parameters to obtain a target recognition model;
the information acquisition module 904 is used for acquiring the action information of the human body to obtain a sample to be detected;
and the target determining module 905 is configured to receive the target identification model sent by the edge server, and perform motion identification on the sample to be detected according to the target identification model to obtain target motion information.
An embodiment of the present application further provides an electronic device, where the electronic device includes: the action recognition method based on the federal edge learning comprises a memory, a processor, a program stored on the memory and capable of running on the processor, and a data bus used for realizing connection communication between the processor and the memory, wherein the program realizes the action recognition method based on the federal edge learning when being executed by the processor. The electronic equipment can be any intelligent terminal including a tablet computer, a vehicle-mounted computer and the like.
Referring to fig. 10, fig. 10 illustrates a hardware structure of an electronic device according to another embodiment, where the electronic device includes:
the processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits, and is configured to execute a relevant program to implement the technical solution provided in the embodiment of the present Application;
the Memory 1002 may be implemented in the form of a Read Only Memory (ROM), a static storage device, a dynamic storage device, or a Random Access Memory (RAM). The memory 1002 may store an operating system and other application programs, and when the technical solution provided by the embodiment of the present disclosure is implemented by software or firmware, the relevant program codes are stored in the memory 1002, and the processor 1001 calls the module management method for executing the online guest system according to the embodiment of the present disclosure;
an input/output interface 1003 for implementing information input and output;
the communication interface 1004 is used for realizing communication interaction between the device and other devices, and can realize communication in a wired mode (such as USB, network cable, and the like) or in a wireless mode (such as mobile network, WI-FI, bluetooth, and the like);
a bus 1005 that transfers information between the various components of the device (e.g., the processor 1001, the memory 1002, the input/output interface 1003, and the communication interface 1004);
wherein the processor 1001, the memory 1002, the input/output interface 1003 and the communication interface 1004 realize communication connections with each other inside the apparatus through a bus 1005.
Furthermore, an embodiment of the present invention also provides a computer-readable storage medium storing computer-executable instructions, which are executed by a processor or a controller, for example, by a processor in the above system embodiment, and can make the processor execute the action recognition method based on federal edge learning in the above embodiment.
The embodiments described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation to the technical solutions provided in the embodiments of the present application, and it is obvious to those skilled in the art that the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems with the evolution of technology and the emergence of new application scenarios.
It will be appreciated by those skilled in the art that the solutions shown in fig. 1-7 are not intended to limit the embodiments of the present application and may include more or fewer steps than those shown, or some of the steps may be combined, or different steps may be included.
The above-described embodiments of the apparatus are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may also be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
One of ordinary skill in the art will appreciate that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.
The terms "first," "second," "third," "fourth," and the like in the description of the application and the above-described figures, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be understood that in the present application, "at least one" means one or more, "a plurality" means two or more. "and/or" for describing an association relationship of associated objects, indicating that there may be three relationships, e.g., "a and/or B" may indicate: only A, only B and both A and B are present, wherein A and B may be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of single item(s) or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is only one type of division of logical functions, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, which are essential or part of the technical solutions contributing to the prior art, or all or part of the technical solutions, may be embodied in the form of a software product stored in a storage medium, which includes multiple instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the methods of the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing programs, such as a usb disk, a portable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The preferred embodiments of the present application have been described above with reference to the accompanying drawings, and the scope of the claims of the embodiments of the present application is not limited thereto. Any modifications, equivalents and improvements that may occur to those skilled in the art without departing from the scope and spirit of the embodiments of the present application are intended to be within the scope of the claims of the embodiments of the present application.
Claims (10)
1. An action recognition method based on federal edge learning is applied to an edge server, the edge server is in communication connection with a plurality of edge devices, and the method is characterized by comprising the following steps:
receiving a training request and equipment information sent by at least one piece of edge equipment, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body;
determining a resource allocation strategy for allocating the edge device according to the device information;
sending the resource allocation strategy and initial model parameters obtained from a pre-configured artificial intelligence model to the edge devices according to the training request, so that each edge device obtains acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters, wherein the acquisition information is obtained by the edge devices according to the resource allocation strategy through acquiring action information of a human body, and the model parameters are obtained by the edge devices according to the initial model parameters through updating;
training the artificial intelligence model according to the plurality of model parameters and the acquired information to obtain a target recognition model;
and sending the target identification model to the edge equipment so that the edge equipment inputs the collected sample to be detected into the target identification model for action identification and outputs target action information of the sample to be detected.
2. The method of claim 1, wherein the device information comprises pilot signals, device parameter information, and device power information; the determining a resource allocation policy for allocating the edge device according to the device information includes:
determining channel state information according to the pilot signal, wherein the channel state information comprises a fading factor;
calculating the fading factor, the equipment parameter information and the equipment power information based on a preset grid search algorithm to obtain the communication transmitting power of the edge equipment;
inputting the communication transmitting power into the edge equipment for calculation to obtain a perception data parameter;
inputting the perception data parameters into a preset training function for calculation to obtain the quantity of perception samples;
and generating the resource allocation strategy for allocating the edge device according to the sensing sample number and the communication transmission power.
3. The method of claim 2, wherein the device power information comprises transmit power and transmit power; the calculating the fading factor, the device parameter information, and the device power information based on the preset grid search algorithm to obtain the communication transmission power of the edge device includes:
calculating a preset exponential integral function, communication bandwidth, noise power spectral density information, the fading factor and the transmission power to obtain a channel capacity function;
and calculating the equipment parameter information, the transmitting power and the channel capacity function according to the grid search algorithm to obtain the communication transmitting power of the edge equipment.
4. The method of claim 2, wherein the training of the artificial intelligence model according to the plurality of model parameters and the collected information to obtain a target recognition model comprises:
carrying out average operation on the plurality of model parameters to obtain global model parameters;
training the artificial intelligence model according to the acquired information to obtain a loss function of the artificial intelligence model;
comparing the loss function of the artificial intelligence model with a preset threshold constraint condition to obtain a comparison result;
when the comparison result is that the function value of the loss function does not meet the threshold constraint condition, sending the global model parameter to the edge device so that the edge device performs multi-step local updating based on the global model parameter until the updated function value of the loss function meets the threshold constraint condition, and obtaining the target identification model according to the updated artificial intelligence model;
and when the comparison result shows that the function value of the loss function meets the threshold constraint condition, obtaining the target recognition model according to the trained artificial intelligence model.
5. An action recognition method based on federal edge learning is applied to edge equipment, an edge server is in communication connection with the edge equipment, and the method is characterized by comprising the following steps:
sending a training request and equipment information to the edge server so that the edge server generates a resource allocation strategy and initial model parameters according to the training request and the equipment information, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body in the environment, and the initial model parameters are used for representing the model performance of an artificial intelligent model in the edge server;
acquiring acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters, wherein the acquisition information is used for representing the acquisition capacity of the acquisition information of the action information of the human body under the resource allocation strategy, and the model parameters are used for representing the updating capacity of the initial model parameters under the initial model parameters;
sending the acquisition information and the plurality of model parameters to the edge server so that the edge server trains the artificial intelligence model according to the acquisition information and the plurality of model parameters to obtain a target recognition model;
collecting motion information of a human body to obtain a sample to be detected;
and receiving the target identification model sent by the edge server, and performing action identification on the sample to be detected according to the target identification model to obtain target action information.
6. The method of claim 5, wherein the resource allocation policy comprises a perceptual sample number; the obtaining of the acquisition information and the plurality of model parameters according to the resource allocation strategy and the initial model parameters includes:
acquiring motion information of a human body according to the number of the perception samples to obtain the acquired information;
and performing multi-step local updating based on the initial model parameters to obtain a plurality of model parameters.
7. The method of claim 5, further comprising:
under the condition that the function value of the loss function of the artificial intelligence model does not meet the threshold constraint condition, receiving the global model parameter sent by the edge server, and performing multi-step local updating based on the global model parameter to obtain an updating result;
and sending the updating result to the edge server until the edge server determines that the function value of the loss function meets the threshold constraint condition according to the updating result.
8. A server, characterized in that the server comprises:
the information receiving module is used for receiving a training request and equipment information sent by at least one piece of edge equipment, wherein the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body;
a policy determining module, configured to determine, according to the device information, a resource allocation policy for allocating the edge device;
a policy sending module, configured to send the resource allocation policy and an initial model parameter obtained from a preconfigured artificial intelligence model to the edge device according to the training request, so that each edge device obtains acquisition information and multiple model parameters according to the resource allocation policy and the initial model parameter, where the acquisition information is obtained by acquiring, by the edge device, action information of a human body according to the resource allocation policy, and the model parameters are obtained by updating, by the edge device, according to the initial model parameter;
the model training module is used for training the artificial intelligence model according to the plurality of model parameters and the acquisition information to obtain a target recognition model;
and the model sending module is used for sending the target identification model to the edge equipment so that the edge equipment inputs the collected sample to be detected into the target identification model for action identification and outputs the target action information of the sample to be detected.
9. A terminal device, characterized in that the terminal device comprises:
the system comprises an information sending module, a resource allocation module and an initial model parameter, wherein the information sending module is used for sending a training request and equipment information to an edge server so as to enable the edge server to generate a resource allocation strategy and the initial model parameter according to the training request and the equipment information, the equipment information is used for representing the acquisition capacity of the edge equipment on the action information of a human body in the environment, and the initial model parameter is used for representing the model performance of an artificial intelligent model in the edge server;
the parameter calculation module is used for obtaining acquisition information and a plurality of model parameters according to the resource allocation strategy and the initial model parameters, wherein the acquisition information is used for representing the acquisition capacity of the acquisition information of the human body under the resource allocation strategy, and the model parameters are used for representing the updating capacity of the initial model parameters under the initial model parameters;
the parameter sending module is used for sending the acquisition information and the plurality of model parameters to the edge server so that the edge server trains the artificial intelligence model according to the acquisition information and the plurality of model parameters to obtain a target recognition model;
the information acquisition module is used for acquiring the action information of the human body to obtain a sample to be detected;
and the target determining module is used for receiving the target identification model sent by the edge server and carrying out action identification on the sample to be detected according to the target identification model to obtain target action information.
10. An electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the federal edge learning based action recognition method of any one of claims 1 to 7 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211474539.1A CN115830709A (en) | 2022-11-23 | 2022-11-23 | Action recognition method based on federal edge learning, server and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211474539.1A CN115830709A (en) | 2022-11-23 | 2022-11-23 | Action recognition method based on federal edge learning, server and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115830709A true CN115830709A (en) | 2023-03-21 |
Family
ID=85530624
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211474539.1A Pending CN115830709A (en) | 2022-11-23 | 2022-11-23 | Action recognition method based on federal edge learning, server and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115830709A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116580448A (en) * | 2023-04-11 | 2023-08-11 | 深圳市大数据研究院 | Behavior prediction method, system, electronic equipment and storage medium |
| CN117763308A (en) * | 2023-12-19 | 2024-03-26 | 深圳市大数据研究院 | Unmanned aerial vehicle-based motion index analysis method, electronic equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111708640A (en) * | 2020-06-23 | 2020-09-25 | 苏州联电能源发展有限公司 | Edge calculation-oriented federal learning method and system |
| WO2021040889A1 (en) * | 2019-08-27 | 2021-03-04 | DawnLight Technologies Inc. | Privacy-preserving activity monitoring systems and methods |
| US20210174243A1 (en) * | 2019-12-06 | 2021-06-10 | International Business Machines Corporation | Efficient private vertical federated learning |
| CN113139663A (en) * | 2021-04-23 | 2021-07-20 | 深圳市大数据研究院 | Federal edge learning configuration information acquisition method, device, equipment and medium |
| CN115174397A (en) * | 2022-07-28 | 2022-10-11 | 河海大学 | Federal edge learning training method and system combining gradient quantization and bandwidth allocation |
| CN115331283A (en) * | 2022-07-25 | 2022-11-11 | 上海中侨健康智能科技有限公司 | Detection system for detecting falling of people in living space and detection method thereof |
| CN115424079A (en) * | 2022-09-30 | 2022-12-02 | 深圳市大数据研究院 | Image classification method based on federal edge learning and related equipment |
-
2022
- 2022-11-23 CN CN202211474539.1A patent/CN115830709A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021040889A1 (en) * | 2019-08-27 | 2021-03-04 | DawnLight Technologies Inc. | Privacy-preserving activity monitoring systems and methods |
| US20210174243A1 (en) * | 2019-12-06 | 2021-06-10 | International Business Machines Corporation | Efficient private vertical federated learning |
| CN111708640A (en) * | 2020-06-23 | 2020-09-25 | 苏州联电能源发展有限公司 | Edge calculation-oriented federal learning method and system |
| CN113139663A (en) * | 2021-04-23 | 2021-07-20 | 深圳市大数据研究院 | Federal edge learning configuration information acquisition method, device, equipment and medium |
| CN115331283A (en) * | 2022-07-25 | 2022-11-11 | 上海中侨健康智能科技有限公司 | Detection system for detecting falling of people in living space and detection method thereof |
| CN115174397A (en) * | 2022-07-28 | 2022-10-11 | 河海大学 | Federal edge learning training method and system combining gradient quantization and bandwidth allocation |
| CN115424079A (en) * | 2022-09-30 | 2022-12-02 | 深圳市大数据研究院 | Image classification method based on federal edge learning and related equipment |
Non-Patent Citations (4)
| Title |
|---|
| KONSTANTIN SOZINOV 等: "Human Activity Recognition Using Federated Learning", 《2018 IEEE INTL CONF ON PARALLEL & DISTRIBUTED PROCESSING WITH APPLICATIONS, UBIQUITOUS COMPUTING & COMMUNICATIONS, BIG DATA & CLOUD COMPUTING, SOCIAL COMPUTING & NETWORKING, SUSTAINABLE COMPUTING & COMMUNICATIONS》, pages 1103 - 1111 * |
| PEIXI LIU 等: "Toward Ambient Intelligence: Federated Edge Learning with Task-Oriented Sensing, Computation, and Communication Integration", 《HTTPS://ARXIV.ORG/PDF/2206.05949V1.PDF》, pages 3 - 4 * |
| 张泽中 等: "面向协同感知的高效通信边缘学习网络架构设计", 《中兴通讯技术》, vol. 28, no. 5, pages 29 - 38 * |
| 高卉杰 等: "人工智能人脸识别技术安全风险评估研究", 《科技与社会》, pages 45 - 53 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116580448A (en) * | 2023-04-11 | 2023-08-11 | 深圳市大数据研究院 | Behavior prediction method, system, electronic equipment and storage medium |
| CN116580448B (en) * | 2023-04-11 | 2024-04-16 | 深圳市大数据研究院 | Behavior prediction method, system, electronic equipment and storage medium |
| CN117763308A (en) * | 2023-12-19 | 2024-03-26 | 深圳市大数据研究院 | Unmanned aerial vehicle-based motion index analysis method, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115830709A (en) | Action recognition method based on federal edge learning, server and electronic equipment | |
| EP3644581B1 (en) | Edge-cloud collaboration system for analyzing internet of things data and operating method thereof | |
| CN110475224B (en) | Sensor data processing and collaborative prediction method based on edge calculation | |
| CN111902996B (en) | Degraded state determination device and degraded state determination method | |
| CN113286315B (en) | Load balance judging method, device, equipment and storage medium | |
| CN116546567B (en) | Data processing method and system based on Bayesian federal learning and electronic equipment | |
| CN109615058A (en) | A kind of training method of neural network model | |
| CN113497647B (en) | Method performed by an electronic device in a communication network and electronic device | |
| CN115065992A (en) | Edge calculation-based cooperative sensing and resource allocation method | |
| CN109598250A (en) | Feature extracting method, device, electronic equipment and computer-readable medium | |
| US20230066703A1 (en) | Method for estimating structural vibration in real time | |
| WO2019062404A1 (en) | Application program processing method and apparatus, storage medium, and electronic device | |
| CN108376099A (en) | A kind of mobile terminal computation migration method of optimization time delay and efficiency | |
| CN109993562B (en) | Satisfaction degree simulation method and device and terminal equipment | |
| CN111291694B (en) | Dish image recognition method and device | |
| CN111723740A (en) | Data identification method, device, equipment and computer readable storage medium | |
| CN111343271A (en) | DTU (data transfer unit) equipment-based artificial intelligence cloud control method and system | |
| CN116560832A (en) | Resource allocation method oriented to federal learning and related equipment | |
| CN112215326A (en) | Distributed AI system | |
| CN112486134B (en) | Multi-object acquisition control method and device and control equipment | |
| CN116010292A (en) | Memory access control method, memory access control device, electronic equipment and storage medium | |
| CN109286408B (en) | Method for energy harvesting and energy receiver | |
| CN114564310A (en) | Data processing method and device, electronic equipment and readable storage medium | |
| CN113676706A (en) | Cooking video generation method and device, server and control system | |
| CN110275809B (en) | Data fluctuation identification method and device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |