Disclosure of Invention
The invention aims to provide a heterogeneous simulation test platform of a joint learning system, which aims to solve the problem that a virtual system heterogeneous test platform similar to a real environment cannot be constructed in the background technology.
In order to achieve the aim, the invention provides the technical scheme that the heterogeneous simulation test platform of the joint learning system comprises a server, a client and an operating environment,
The server provides computing service as a node of the network, and stores and processes data and information on the network;
The client corresponds to the server, and a program for providing local service for the client is generally installed on a common client and needs to be matched with the server for running;
and the running environment judges the simulation parameters of the virtual environment.
Preferably, the server may declare various parameters of the client, which send data and information to the client via an initialized client list/parameters, load data (detection), load model, and learn.
Preferably, the client sequentially performs environment initialization, load data and load model, the environment initialization data and information are sent to the operation environment, and the initial values of each round of operation environment are the same and normally distributed.
Preferably, the virtual simulation environment parameters in the operating environment include operating speed, net speed, communications and hardware.
Preferably, the loading data (detection) distributes the data set to the loading data of the client, and the data and information of the loading data are sent to the loading model.
Preferably, the data and information of the load model are split into two lines, one for local learning and one for local testing.
Preferably, the data and information of the loading model are sent to learning to judge, when judging as 'Y', the model is distributed to local learning, and when judging as 'N', the data and information are sent to global test.
Preferably, the locally learned data and information is sent to the aggregate after loss/accuracy/model parameters, and the aggregate data and information is sent to the global test.
Preferably, the global test sends data and information to the local test after passing through the global model.
Preferably, the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And secondly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step three, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Compared with the prior art, the invention has the beneficial effects that:
The heterogeneous simulation test platform of the joint learning system performs abstraction and modeling according to a known system, uses simulation and abstraction modes to simulate and abstract the problem of limitation of hardware, network, computing resources and the like in a joint learning real environment, solves the heterogeneous problem of the joint learning system, performs abstraction and modeling according to known system data such as hardware computing capacity, network delay, communication bandwidth and the like in an actual joint learning application scene, constructs a virtual system heterogeneous test platform similar to the real environment, can simulate different actual application scenes to verify different joint learning solutions, has stronger practicability, and realizes combination and sharing of data information resources, hardware equipment resources and manpower resources among different databases. The key point is that a global data mode or a global external view is established based on a local database mode, meanwhile, collected data also supports access to historical data, a user performs decision-supported query through a unified data interface provided by a data warehouse, the global mode is especially important for establishing an advanced decision support system, a joint learning system can collect parameters of a model, a server coordinates edge devices to participate in learning, each edge device has learning data, each edge device learns a local model by utilizing own data, the own parameters are encrypted or not encrypted and uploaded to the server, the server performs average or weighted average on the collected parameters and broadcasts the parameters to each edge device, the joint learning can generate a more intelligent model, lower delay and lower power consumption, and simultaneously, the privacy of the user is ensured, the cloud forms collaborative update for a shared model, unhook the demand of machine learning and cloud storage data, make the model more smart, lower in delay and more energy-saving, protect the privacy of users from being threatened, besides realizing the update of the shared model, the users can use the improved model immediately, the obtained experience can be different according to different personal use modes, relevant data is accessed through the internet of things, then model learning, model updating and calculation storage are carried out locally, a series of aggregation calculation and processing are carried out on the models provided by each user, and the combined global model is issued to each user, iterated back and forth until a better model is learned, thereby being convenient for the users to call better and share value, being applicable to edge devices with different calculation, communication and storage capacities, having stronger functionality and practicability, the operation is convenient, and the combined learning effect is better.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention provides a technical scheme that the heterogeneous simulation test platform of a joint learning system comprises a server, a client and an operating environment,
The server provides computing service as a node of the network, and stores and processes data and information on the network;
The client corresponds to the server, and a program for providing local service for the client is generally installed on a common client and needs to be matched with the server for running;
The running environment judges the simulation parameters of the virtual environment.
Further, the server may declare various parameters of the client, load data (detection), load model, and learn, declare various parameters of the client to send data and information to the client via an initialized client list/parameters.
Further, the client sequentially performs environment initialization, load data and load models, the data and information of the environment initialization are sent to the operation environment, and the initial values of each round of operation environment are the same and normally distributed.
Further, virtual simulation environment parameters in the operating environment include operating speed, net speed, communications, and hardware.
Further, the loading data (detection) distributes the data set to the loading data of the client, and the data and information of the loading data are sent to the loading model.
Further, the data and information of the load model are split into two lines, one to local learning and one to local testing.
Further, data and information of the loaded model are sent to learning to be judged, when the judgment is "Y", the model is distributed to local learning, and when the judgment is "N", the data and the information are sent to global test.
Further, the locally learned data and information is sent to the aggregate after loss/accuracy/model parameters, and the aggregate data and information is sent to the global test.
Further, the global test sends data and information to the local test after passing through the global model.
Embodiment one:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And secondly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step three, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Embodiment two:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And step two, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And thirdly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Embodiment III:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And secondly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step three, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, the locally learned data and information are transmitted to the total after loss/accuracy/model parameters, and the total data and information are transmitted to the global test.
And seventhly, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Embodiment four:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And step two, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And thirdly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, the locally learned data and information are transmitted to the total after loss/accuracy/model parameters, and the total data and information are transmitted to the global test.
And seventhly, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Fifth embodiment:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And secondly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And thirdly, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And step four, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Example six:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And secondly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And thirdly, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fourthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And fifthly, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Embodiment seven:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
And secondly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And thirdly, dividing the data and information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Example eight:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
the server can declare various parameters of the client, load data (detection), load model and study, and the various parameters of the client can send data and information to the client through the initialized client list/parameters.
The virtual simulation environment parameters in the operation environment comprise operation speed, net speed, communication and hardware
And thirdly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed. .
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Example nine:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
The virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And secondly, the server can declare various parameters of the client, load data (detection), load a model and learn, and the various parameters of the client can be declared to send data and information to the client through an initialized client list/parameter.
And thirdly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And seventhly, sending the locally learned data and information to the total after loss/accuracy/model parameters, and sending the total data and information to the global test.
And step eight, the global test sends the data and the information to the local test after passing through the global model.
Example ten:
the heterogeneous simulation test platform of the joint learning system comprises the following steps:
The virtual simulation environment parameters in the running environment comprise running speed, net speed, communication and hardware.
And secondly, the server can declare various parameters of the client, load data (detection), load a model and learn, and the various parameters of the client can be declared to send data and information to the client through an initialized client list/parameter.
And thirdly, sequentially carrying out environment initialization, load data and load models by the client, and sending the data and information of the environment initialization to the operation environment, wherein the initial values of each round of operation environment are the same, and the data and the information are normally distributed.
And step four, loading data (detecting) and distributing the data set to load data of the client, and sending the data and information of the load data to a load model.
And fifthly, dividing the data and the information of the load model into two lines, wherein one line is sent to local learning and the other line is sent to local testing.
And step six, data and information of the loaded model are sent to learning to be judged, when the judgment is Y, the model is distributed to local learning, and when the judgment is N, the data and the information are sent to global test.
And step seven, the global test sends the data and the information to the local test after passing through the global model.
And step eight, the locally learned data and information are sent to the aggregate after loss/accuracy/model parameters, and the aggregate data and information are sent to the global test.
Finally, it should be noted that the above description is only for illustrating the technical solution of the present invention, and not for limiting the scope of the present invention, and that the simple modification and equivalent substitution of the technical solution of the present invention can be made by those skilled in the art without departing from the spirit and scope of the technical solution of the present invention.