CN114727313A

CN114727313A - Information processing method, device, equipment and storage medium

Info

Publication number: CN114727313A
Application number: CN202110008074.XA
Authority: CN
Inventors: 邓娟; 王莹莹; 李娜; 孙军帅; 刘光毅
Original assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Current assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Priority date: 2021-01-05
Filing date: 2021-01-05
Publication date: 2022-07-08

Abstract

The invention discloses an information processing method, an information processing device, information processing equipment and a storage medium. Wherein the method comprises the following steps: receiving first information respectively sent by at least two second nodes; the first information represents relevant parameters of a machine learning model trained by the second node; for each second node, judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and notifying the corresponding second node of the first learning mode; wherein the first learning mode is used for the corresponding second node to retrain the machine learning model so that the performance of the machine learning model meets a preset condition.

Description

Information processing method, device, equipment and storage medium

Technical Field

The present invention relates to the field of wireless network technologies, and in particular, to an information processing method, apparatus, device, and storage medium.

Background

With the rapid development of communication network technology and artificial intelligence technology, the network element equipment can train a machine learning model, and predict the performance index of the stored network data by using the trained machine learning model. Generally, in daily maintenance of the network element device, such as cutting and restarting, a machine learning model trained by the network element device may not be applicable, so that accuracy of a performance index of network data predicted by the network element device is reduced.

Therefore, it is desirable to find a technical solution for improving the accuracy of the performance index of the network data predicted by the network element device.

Disclosure of Invention

In view of the above, embodiments of the present invention are intended to provide an information processing method, apparatus, device, and storage medium.

The technical scheme of the embodiment of the invention is realized as follows:

at least one embodiment of the present invention provides an information processing method, including:

receiving first information respectively sent by at least two second nodes; the first information represents relevant parameters of a machine learning model trained by the second node;

for each second node, judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information;

when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and notifying the corresponding second node of the first learning mode;

wherein the first learning mode is used for the corresponding second node to retrain the machine learning model so that the performance of the machine learning model meets a preset condition.

Further in accordance with at least one embodiment of the present invention, the first information includes a second learning mode in which the second node trains a machine learning model; the determining a first learning mode corresponding to the respective second node comprises:

for other nodes except the corresponding second node in the at least two second nodes, judging whether a second learning mode of the machine learning model trained by the other nodes is a transfer learning mode or not by using the first information;

when the second learning mode of the machine learning model trained by the other nodes is determined to be a transfer learning mode, taking the other nodes as nodes to be processed to obtain a plurality of nodes to be processed;

evaluating the performance of the machine learning model trained by the nodes to be processed to obtain an evaluation result;

determining a machine learning model with optimal performance evaluation in the evaluation result; taking the transfer learning mode corresponding to the machine learning model with the optimal performance evaluation as the first learning mode corresponding to the corresponding second node;

wherein the parameters of the transfer learning mode characterization machine learning model are obtained through transfer learning.

Further in accordance with at least one embodiment of the present invention, the notifying the respective second nodes of the first learning mode includes:

generating second information; the second information comprises the name or number of the corresponding second node, a recommended first learning mode, a migrated source machine learning model, and the name or number of the node corresponding to the migrated source machine learning model;

and sending the second information to the corresponding second node.

for other nodes except the corresponding second node in the at least two second nodes, judging whether second learning modes of the machine learning models trained by the other nodes are not transfer learning modes by using the first information;

when the learning mode of the machine learning model trained by other nodes is determined not to be the transfer learning mode, taking the autonomous learning mode as the first learning mode corresponding to the corresponding second node;

wherein the autonomous learning mode characterizes the respective second node to keep model parameters unchanged while retraining the machine learning model.

generating third information; the third information comprises the name or number of the corresponding second node, and a recommended first learning mode;

and sending the third information to the corresponding second node.

for other nodes except the corresponding second node in the at least two second nodes, judging whether the learning mode of the machine learning model trained by the other nodes is a transfer learning mode or not by using the first information;

when the learning mode of the machine learning model trained by the other nodes is determined to be a transfer learning mode, taking the other nodes as nodes to be processed to obtain a plurality of nodes to be processed;

taking a combination of a transfer learning mode and an autonomous learning mode corresponding to the machine learning model with the performance evaluation larger than the performance threshold value in the evaluation result as a first learning mode corresponding to the corresponding second node;

and the combination of the transfer learning mode and the autonomous learning mode represents that the corresponding second node retrains the machine learning model respectively under the condition of keeping the model parameters unchanged and under the condition of updating the model parameters.

Further in accordance with at least one embodiment of the present invention, the sending to the respective second nodes includes:

generating fourth information; the fourth information comprises the name or number of the corresponding second node, a recommended first learning mode, a migrated source machine learning model, and the name or number of the node corresponding to the migrated source machine learning model;

and sending the fourth information to the corresponding second node.

At least one embodiment of the present invention provides an information processing method applied to a second node, the method including:

reasoning the predicted value output by the trained machine learning model to obtain a reasoning result;

evaluating the reasoning result to obtain an evaluation result; judging whether to report the evaluation result to the first node or not; when the evaluation result is determined to be reported to the first node, generating first information; the first information represents relevant parameters of a machine learning model trained by the second node; and sending the first information to the first node.

Further, in accordance with at least one embodiment of the present invention, the generating the first information includes:

and generating first information based on the name or the number of the second node and the learning mode and the parameters of the machine learning model trained by the second node.

Further, in accordance with at least one embodiment of the present invention, the method further comprises:

receiving a first learning mode sent by the first node; the first learning mode is determined by the first node when the performance of the machine learning model trained by the second node is determined to be not in accordance with a preset condition; retraining the machine learning model by using the first learning mode so as to enable the performance of the machine learning model to accord with a preset condition.

notifying a third node of the first learning mode;

the first learning mode is used for the third node to retrain the machine learning model so that the performance of the machine learning model meets a preset condition.

At least one embodiment of the present invention provides an information processing apparatus including:

the receiving unit is used for receiving first information respectively sent by at least two second nodes; the first information represents relevant parameters of a machine learning model trained by the second node;

the first processing unit is used for judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not according to the first information for each second node; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and informing the corresponding second node of the first learning mode;

the second processing unit is used for reasoning the predicted value output by the trained machine learning model to obtain a reasoning result; evaluating the reasoning result to obtain an evaluation result;

the third processing unit is used for judging whether to report the evaluation result to the first node or not; when the evaluation result is determined to be reported to the first node, generating first information; the first information represents relevant parameters of a machine learning model trained by the second node;

a sending unit, configured to send the first information to the first node.

At least one embodiment of the present invention provides a first node apparatus including:

the first communication interface is used for receiving first information respectively sent by at least two second nodes; the first information represents relevant parameters of a machine learning model trained by the second node;

the first processor is used for judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not according to the first information for each second node; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and informing the corresponding second node of the first learning mode;

At least one embodiment of the present invention provides a second node apparatus including:

the second processor is used for reasoning the predicted value output by the trained machine learning model to obtain a reasoning result; evaluating the reasoning result to obtain an evaluation result; judging whether to report the evaluation result to the first node or not; when the evaluation result is determined to be reported to the first node, generating first information; the first information represents relevant parameters of a machine learning model trained by the second node;

a second communication interface for sending the first information to the first node.

At least one embodiment of the invention provides a first node device comprising a processor and a memory for storing a computer program capable of running on the processor,

wherein the processor is configured to execute the steps of the method of any one of the first node device sides when running the computer program.

At least one embodiment of the invention provides a second node device, comprising a processor and a memory for storing a computer program capable of running on the processor,

wherein the processor is configured to, when running the computer program, perform the steps of the method at any of the second node device sides.

At least one embodiment of the invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of any of the methods described above.

The information processing method, the device, the equipment and the storage medium provided by the embodiment of the invention receive first information respectively sent by at least two second nodes; the first information represents relevant parameters of a machine learning model trained by the second node; for each second node, judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and notifying the corresponding second node of the first learning mode; wherein the first learning mode is used for the corresponding second node to retrain the machine learning model so that the performance of the machine learning model meets a preset condition. By adopting the technical scheme of the embodiment of the invention, the first node can judge whether the second node needs to update the machine learning model or not by utilizing the first information reported by the second node, and sends the corresponding first learning mode to the second node when the second node is determined to need to update the machine learning mode, so that the second node can update the machine learning model by utilizing the first learning mode to predict the performance index of the network data by utilizing the updated machine learning model, thereby improving the accuracy of the predicted performance index of the network data.

Drawings

FIG. 1 is a first schematic flow chart of an implementation of an information processing method according to an embodiment of the present invention;

FIG. 2 is a first flowchart illustrating an implementation of interaction between a first node and a second node according to an embodiment of the present invention;

FIG. 3 is a schematic diagram illustrating an implementation flow of interaction between a first node and a second node according to an embodiment of the present invention;

FIG. 4 is a third schematic flow chart illustrating an implementation of interaction between a first node and a second node according to an embodiment of the present invention;

fig. 5 is a schematic diagram of physical entities corresponding to a first node and a second node according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of a second implementation flow of the information processing method according to the embodiment of the present invention;

fig. 7a and 7b are schematic diagrams of modules corresponding to a first node, a second node, and a third node according to an embodiment of the present invention;

FIG. 8 is a fourth flowchart illustrating an implementation of interaction between a first node and a second node according to an embodiment of the present invention;

FIG. 9 is a first block diagram of an information processing apparatus according to an embodiment of the present invention;

FIG. 10 is a second schematic diagram of an information processing apparatus according to an embodiment of the present invention;

FIG. 11 is a block diagram of an information processing system according to an embodiment of the present invention;

fig. 12 is a first schematic structural diagram of a first node device according to an embodiment of the present invention;

fig. 13 is a schematic structural diagram of a second node device according to an embodiment of the present invention.

Detailed Description

Before the technical solution of the embodiment of the present invention is introduced, a description is given of a related art.

In the related art, at present, a large number of homogeneous network elements exist in a mobile communication network, where the homogeneous network elements may refer to network elements with the same function, load sharing, and strong correlation of network performance indexes, for example, network elements in the same resource Pool (Pool) in a core network are homogeneous network elements, that is, network elements with the same performance index set, shared user number and load, and very similar performance measurement data in most cases. Here, the homogeneous network element may specifically be a group of base stations in the wireless network that exhibit similar wireless performance data due to the deployment scenario approximation.

Generally, for homogeneous network elements, in order to reduce training workload, the same Artificial Intelligence (AI) model is often used to predict the index performance of network data generated during daily network operation and maintenance work, and report the performance to a server of a network management platform. During special maintenance work of an individual network element, such as cutting, restarting, etc., the AI model trained by the network element may not be applicable, and thus the performance of the network element within the same pool may also be affected. In addition, because the storage capacity of the server is limited, each network element can only train the AI model by using data stored for 1 month at most, for example, according to the granularity of 15min of the data acquisition period, each network performance index can only extract 2880 sample points, which is 96 × 30 at a time, and the training of the AI model is very limited, so that the accuracy of the AI model is affected.

In summary, in the daily maintenance of the network element device, such as cutting, restarting, etc., the machine learning model trained by the network element device may not be applicable, so that the accuracy of the performance index of the network data predicted by the network element device is reduced. In addition, the storage capacity of the network management server is limited, only 1 month of data is stored at most once, and the AI model training effect is influenced; if the same set of AI model is adopted among the homogeneous network elements, the performance change of the individual network elements during the operation and maintenance engineering can not be dealt with; if the models are trained separately, the training workload is huge. In the practical AI application process, data acquisition consumes processing resources and transmission resources greatly, so it is of practical value if the data acquisition frequency can be reduced and even if an AI model with a satisfactory performance can be trained without acquiring data. Meanwhile, the maximum storage space of each node is limited, when a plurality of AI training tasks exist, the storage space is bound to be tight, if the consumption of the storage space can be reduced as much as possible, the number of models which can be trained by the nodes at the same time is increased, and the intelligent generation capacity of the nodes is improved.

Based on this, in each embodiment of the present invention, first information respectively sent by at least two second nodes is received; the first information represents relevant parameters of a machine learning model trained by the second node; for each second node, judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and notifying the corresponding second node of the first learning mode; wherein the first learning mode is used for the corresponding second node to retrain the machine learning model so that the performance of the machine learning model meets a preset condition.

It should be noted that, in the embodiment of the present invention, the AI model training framework based on the transfer learning has the following advantages:

on the premise of ensuring the model effect, when the network element equipment changes the index performance of the network data by using the trained machine learning model due to factors such as operation and maintenance operations on certain network element equipment, the first node can timely update the machine learning model of the second node, namely the network element equipment on line through the conversion of the first learning mode, so that the performance index of the network data is predicted by using the updated machine learning model, and the influence on the abnormal detection of the performance index is reduced.

The present invention will be described in further detail with reference to the drawings and examples.

An embodiment of the present invention provides an information processing method, which is applied to a first node, and as shown in fig. 1, the method includes:

step 101: receiving first information respectively sent by at least two second nodes; the first information represents relevant parameters of a machine learning model trained by the second node;

step 102: for each second node, judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information; and when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node.

Step 103: notifying the respective second node of the first learning mode;

Here, in step 101, in actual application, the first information may be an evaluation result obtained by the second node evaluating and predicting the index performance of the network data by using the trained machine learning model, and specifically, the second node may infer a predicted value output by the trained machine learning model to obtain an inference result, and evaluate the inference result to obtain an evaluation result. After the second node reports the first information to the first node, the first node may determine, by using the first information, whether the machine learning model currently trained by the second node needs to be updated, so that the second node can update the machine learning model currently trained and predict the index performance of the network data by using the updated machine learning model, thereby ensuring the accuracy of the prediction result.

Here, in step 102, in practical application, the at least two second nodes may specifically be homogeneous network elements. When the first node determines that a certain second node needs to update the trained machine learning model by using the first information, a first learning mode corresponding to the second node is determined and notified to the second node, and the second node can retrain the machine learning model by using the first learning mode, so that the result of predicting the index performance of the network data by using the machine learning model is more accurate.

The following describes in detail how the first node instructs the corresponding second node to update the trained machine learning model.

In a first case, when the first node determines that a certain second node needs to update the trained machine learning model, the first node determines a first learning mode used by the second node to retrain the machine learning model by using first information reported by other nodes except the second node in the at least two nodes.

In practical application, after the at least two second nodes report respective first information to the first node, the first node may determine, by using the first information, whether performance of a machine learning model trained by the second node meets a preset condition, that is, whether a result of predicting index performance of network data by using the trained machine learning model is within a preset range, and when it is determined that performance of the machine learning model trained by a certain second node does not meet the preset condition, a learning mode used by the machine learning model trained by another node may be updated by using a learning mode used by the machine learning model trained by another node.

Based on this, in an embodiment, the first information includes a second learning mode in which the second node trains a machine learning model; the determining a first learning mode corresponding to the respective second node comprises:

for other nodes except the corresponding second node in the at least two second nodes, judging whether a second learning mode of a machine learning model trained by the other nodes is a transfer learning mode or not by using the first information;

Here, the first information may further include an index performance prediction result obtained by predicting the index performance of the network data by the second node using the machine learning model, so that the first node may compare the index performance prediction results respectively corresponding to the plurality of nodes to be processed with a preset index threshold to obtain comparison results, and select, from the comparison results, a node corresponding to the machine learning model with the performance evaluation being optimal by the node whose performance index performance prediction result is closest to the preset index threshold.

For example, table 1 is an illustration of first information reported to a first node by a second node, as shown in table 1, if the first node determines that the performance of a machine learning model trained by the first node does not meet a preset condition by using the first information reported by a second node 5, the first node may determine that the learning modes of the second node 1, the second node 2, and the second node 3 are migration learning modes by using the first information, and use the second node 1, the second node 2, and the second node 3 as nodes to be processed, and by using an index performance prediction result in the first information and combining a preset index threshold, the performance of the machine learning models trained by the second node 1, the second node 2, and the second node 3 respectively can be evaluated and ranked to obtain a node corresponding to the machine learning model with the optimal performance evaluation, and as shown in table 2, assuming that the first node is the second node 1, the second node 5 is informed of the migration learning mode corresponding to the second node 1.

Second node numbering	First information
		Second node 1	Transition learning mode, index performance prediction result 1
Second node 2	Migration learning mode, index performance prediction result 2
		Second node 3	Migration learning mode, index performance prediction result 3
Second node 4	Autonomous learning mode, index performance prediction result 4
		Second node 5	Autonomous learning mode, index performance prediction result 5

TABLE 1

Second node number	Performance evaluation ranking
		Second node 1	Performance evaluation optimization
Second node 2	Performance evaluation second
		Second node 3	Performance evaluation ofIII

TABLE 2

In practical application, after the at least two second nodes report respective first information to the first node, the first information may be used to determine a second node that needs to retrain the machine learning model, and the first information reported by other nodes except the second node in the at least two second nodes is used to determine a learning mode used by the second node to retrain the machine learning model.

Based on this, in an embodiment, the notifying the respective second node of the first learning mode includes:

and sending the second information to the corresponding second node.

In an example, taking a first node as a CU and a second node as a DU as an example, an interaction process between the first node and the second node is described, as shown in fig. 2, including:

step 201: the CU receives the first information sent by DU-A and DU-B respectively.

Here, the CU is used as the first node, and is configured to aggregate and evaluate first information reported by all distributed DU nodes, that is, parameters of a machine learning model trained by the DU nodes and an index performance evaluation result obtained by predicting index performance of network data by using the corresponding machine learning model; and generates a new learning mode proposal for each DU node.

Step 202: for DU-A, a CU judges whether the performance of a machine learning model trained by DU-A meets a preset condition or not; when it is determined that the performance of the DU-a trained machine learning model does not meet the preset condition, step 203 is emutexecuted.

Here, when the CU determines that the performance of the DU-a trained machine learning model does not meet the preset condition, it may determine whether the learning mode of the DU-B trained machine learning model is the transition learning mode, and if the learning mode of the DU-B trained machine learning model is the transition learning mode, recommend the learning mode used by the DU-B trained machine learning model to the DU-a for use.

Step 203: determining a first learning mode corresponding to DU-A as a transfer learning mode; generating second information; and sending the second information to DU-A.

Here, the second information may specifically be: the content of the Learning mode information may include the name or number of the DU-a node, the recommended Learning mode is a migration Learning mode, a migration source machine Learning model, the name/number of the node corresponding to the source machine Learning model, the number of the source machine Learning model, and so on. Wherein the migrated source machine learning model is a DU-B trained machine learning model, e.g., 3 × 3, W, D, where 3 × 3 denotes that the DU-B trained machine learning model includes 3 layers, W denotes the weight of the model, and D denotes the weight of the model.

Here, after receiving the Learning mode instruction message, the DU-a node may update the Learning mode according to the instruction, and retrain the machine Learning model according to the new Learning mode; the migrated source machine learning model and its corresponding node name/number, the source machine learning model number, etc. may also be saved.

Here, the procedure for DU-B is similar to that for DU-A and is not described in detail here.

Here, the CU as a functional entity may implement the following functions:

1. and evaluating the performance of the machine learning model reported by each DU node according to the standard reaching condition, if the machine learning model trained by a certain DU node does not reach the standard, generating a new learning mode suggestion for the DU node, and sending the new learning mode suggestion to the corresponding DU node through a learning mode instruction message.

2. The content of the Learning mode information message may include the name or number of the node, a recommended Learning mode for the node (including migration Learning), a source machine Learning model of the migration (including one machine Learning model), the node name/number corresponding to the source machine Learning model, and the number of the source machine Learning model.

3. Evaluating the performance of the machine learning model reported by each DU node according to the standard reaching condition, if the machine learning model trained by a certain DU node reaches the standard and a plurality of machine learning models exist, selecting a recommended model for the DU node according to the strategy in all the standard reaching machine learning models, sending the recommended model to the corresponding DU node through an Online model instruction message, after receiving the Online model instruction message, using the model indicated in the message to perform Online reasoning by an Online reasoning module of the DU node, if the message comprises the generation time of the machine learning model, judging whether the model fails by using the generation time of the machine learning model, and if the model fails, adopting the model; if the online reasoning model fails, selecting the model as the online reasoning model; wherein, the online reasoning module of the DU node can also delete other unused models.

4. The content of the Online model instruction message may include a node name, a number of the machine learning model, and a generation time of the machine learning model.

5. The model performance standard-reaching judgment condition may be configured, for example, whether the machine learning model trained by the DU node meets the standard is judged according to a preset performance threshold.

6. The model selection strategy may be configured, for example, to select a best performing machine learning model as the migration source machine learning model, or to train a node with the least data storage as the migration source machine learning model, and so on.

In the second case, when the first node determines that a second node needs to update the trained machine learning model, the first node may be used to train the machine learning model again using the first learning mode used by the second node.

In practical application, after the at least two second nodes report respective first information to the first node, the first node may determine, by using the first information, whether performance of a machine learning model trained by the second node meets a preset condition, that is, whether a result of predicting index performance of network data by using the trained machine learning model is within a preset threshold range, and when it is determined that performance of the machine learning model trained by a certain second node does not meet the preset condition, the first node may train a learning mode used by the machine learning model trained by the second node again.

for other nodes except the corresponding second node in the at least two second nodes, judging whether second learning modes of machine learning models trained by the other nodes are not all transfer learning modes by using the first information;

For example, table 3 is an illustration of first information reported to the first node by the second node, and as shown in table 3, assuming that the first node determines that the performance of the machine learning model trained by the node does not meet the preset condition by using the first information reported by the second node 5, and determines that the learning mode of the second node 1, the second node 2, the second node 3, and the second node 4 is not the transfer learning mode, the autonomous learning mode is notified to the second node 5.

Second node numbering	First information
		Second node 1	Autonomous learning mode, index performance prediction result 1
Second node 2	Autonomous learning mode, index performance prediction result 2
		Second node 3	Autonomous learning mode, index performance prediction result 3
Second node 4	Autonomous learning mode, index performance prediction result 4
		Second node 5	Autonomous learning mode, index performance prediction result 5

TABLE 3

In practical application, after the at least two second nodes report respective first information to the first node, the first node may determine, by using the first information, a second node that needs to retrain the machine learning model, and determine that a learning mode used by the second node to retrain the machine learning model is an autonomous learning mode.

and sending the third information to the corresponding second node.

Here, the recommended first learning mode is an autonomous learning mode, that is, the corresponding second node keeps the model parameters unchanged when the machine learning model is retrained, for example, the machine learning model is retrained again by using the reselected network data.

In an example, taking a first node as a CU and a second node as a DU as an example, an interaction process between the first node and the second node is described, as shown in fig. 3, including:

step 301: the CU receives the first information sent by DU-A and DU-B respectively.

Step 302: for DU-A, a CU judges whether the performance of a machine learning model trained by DU-A meets a preset condition or not; when it is determined that the performance of the DU-a trained machine learning model does not meet the preset condition, step 303 is performed.

Here, when the CU determines that the performance of the DU-a trained machine learning model does not meet the preset condition, it may determine whether the learning mode of the DU-B trained machine learning model is the transition learning mode, and set the learning mode used by the DU-a retrained machine learning model to the autonomous learning mode, assuming that the learning mode of the DU-B trained machine learning model is not the transition learning mode.

Step 303: determining a first learning mode corresponding to DU-A as an autonomous learning mode; generating third information; and sending the third information to DU-A.

Here, the third information may specifically be: the Learning mode instruction message, the content of which may include the name or number of the DU-a node, and the recommended Learning mode being an autonomous Learning mode.

Here, after receiving the Learning mode instruction message, the DU-a node may update the Learning mode according to the instruction, and retrain the machine Learning model according to the autonomous Learning mode.

Here, the functional entity corresponding to the first node interacts with the functional entity corresponding to the second node, and the following advantages are provided:

(1) the model effect is guaranteed, and meanwhile, the data reporting amount or the storage space requirement is reduced; or the performance of the model can be improved on the premise of a certain total amount of data storage space.

(2) Through the conversion of the learning mode, when the index model changes due to the operation and maintenance operation on a certain network element, the online models of all the network elements can be updated in time, and the influence on the model performance is reduced.

In a third case, when the first node determines that a certain second node needs to update the trained machine learning model, the first node determines, by using first information reported by other nodes except the second node in the at least two nodes, a first learning mode used by the second node to retrain the machine learning model.

In practical application, after the at least two second nodes report respective first information to the first node, the first node may determine, by using the first information, whether performance of a machine learning model trained by the second node meets a preset condition, that is, whether a result of predicting index performance of network data by using the trained machine learning model is within a preset range, and when it is determined that performance of the machine learning model trained by a certain second node does not meet the preset condition, the learning mode used by the machine learning model trained by another node and the learning mode used by the machine learning model trained by the second stage may be used to update the learning mode used by the machine learning model trained by the second node.

and the combination of the transfer learning mode and the autonomous learning mode represents that the corresponding second node retrains the machine learning model under the condition of keeping the model parameters unchanged and under the condition of updating the model parameters.

For example, table 4 is an illustration of first information reported to the first node by the second node, as shown in table 3, it is assumed that the first node determines, by using the first information reported by the second node 5, that the performance of the machine learning model trained by the node does not meet the preset condition, and determines that neither the learning mode of the second node 1 nor the learning mode of the second node 4 is the migration learning mode, and the learning modes of the second node 2 and the second node 3 are the migration learning modes, and if the index performance prediction results respectively corresponding to the first node 2 and the second node 3 are greater than the performance threshold, the migration learning mode used by the machine learning model trained by the second node 2 and the migration learning mode used by the machine learning model trained by the second node 3 are combined with the autonomous learning mode, and the combined learning mode is notified to the second node 5 for the second node 5 to perform the recommended learning mode, retraining its own machine learning model.

Second node numbering	First information
		Second node 1	Autonomous learning mode, index performance prediction result 1
Second node 2	Migration learning mode, index performance prediction result 2
		Second node 3	Migration learning mode, index performance prediction result 3
Second node 4	Autonomous learning mode, index performance prediction result 4
		Second node 5	Autonomous learning mode, index performance prediction result 5

TABLE 4

In practical application, after the at least two second nodes report respective first information to the first node, the first node may determine, by using the first information, a second node that needs to retrain the machine learning model, and determine that a learning mode used by the second node to retrain the machine learning model is a combination of an autonomous learning mode and a transfer learning mode.

In practical application, the sending to the corresponding second node includes:

and sending the fourth information to the corresponding second node.

Here, the recommended first learning mode is a combination of an autonomous learning mode and a transition learning mode, that is, the corresponding second node retrains the machine learning model while keeping the model parameters unchanged, and performs retraining after updating the machine learning model with the indicated model parameters of other second nodes.

In an example, taking a first node as a CU and a second node as a DU as an example, an interaction process between the first node and the second node is described, as shown in fig. 4, including:

step 401: the CU receives the first information sent by DU-A, DU-B and DU-C, respectively.

Step 402: for DU-A, a CU judges whether the performance of a machine learning model trained by DU-A meets a preset condition or not; when it is determined that the performance of the DU-a trained machine learning model does not meet the preset condition, step 403 is performed.

Here, when the CU determines that the performance of the DU-a trained machine learning model does not meet the preset condition, it may determine whether the learning modes of the DU-B and DU-C trained machine learning models are both the migration learning mode, and if the learning modes of the DU-B and DU-C trained machine learning models are both the migration learning mode and the indemutemutex performance prediction results reported by the DU-B and DU-C are both greater than the performance threshold, combine the migration learning mode of the DU-B, the migration learning mode of the DU-C, and the autonomous learning mode, and set the learning mode used by the DU-a retrained machine learning model as the combination of the migration learning mode of the DU-B, the migration learning mode of the DU-C, and the autonomous learning mode.

Step 403: determining a first learning mode corresponding to DU-A as a combination of an autonomous learning mode and a transfer learning mode; and generating fourth information and transmitting the fourth information to the DU-a.

Here, the fourth information may specifically be: the Learning mode information may include the name or number of the DU-a node, and the recommended Learning mode may be a combination of an autonomous Learning mode and a transition Learning mode.

Here, the procedure for DU-B, DU-C is similar to that for DU-A and is not described in detail here.

The following describes in detail the implementation process of the information processing method according to the embodiment of the present invention with reference to specific embodiments.

Fig. 5 is a schematic diagram of physical entities corresponding to a first node and a second node in the embodiment of the present invention, and as shown in fig. 5, the system includes:

and the two base stations are respectively represented by DU A and DU B and are used for sending the relevant parameters of the self-trained machine learning model, such as the performance index evaluation result, to the CU.

A central facility base station, denoted CU, may comprise a "model evaluation" entity for evaluating the performance of the machine learning model trained by the respective DU.

Specifically, for DU B, if the model reaches the standard, selecting a recommended model for the node according to the strategy in all the models reaching the standard, and indicating the model to the B base station through the message;

for DU A, the initial learning mode is autonomous learning, and if the initial learning mode is tracked to have a stable standard model in a longer time, the learning mode is changed into autonomous + transfer learning (the source model is the model of DU B); at this time, the DU A generates a new transfer learning model, and sends an evaluation result to a model evaluation entity corresponding to the CU, and if the performance of the transfer learning model reaches the standard, the DU A can be instructed to adopt transfer learning;

due to the dynamic change of the scene, the performance of the migration learning model adopted by the DU A is reduced and does not reach the standard any more. The model evaluation entity corresponding to the CU indicates to change the learning mode of the CU into autonomous + transfer learning; at this time, the DU a generates a new autonomous learning model, and sends the evaluation result to the model evaluation entity, and if the performance of the autonomous learning model reaches the standard, the DU a may be instructed to adopt autonomous learning.

By adopting the technical scheme provided by the embodiment of the invention, the first node can judge whether the second node needs to update the machine learning model or not by using the first information reported by the second node, and when the second node is determined to need to update the machine learning model, the corresponding first learning mode is sent to the second node, so that the second node can update the machine learning model by using the first learning mode, and the performance index of the network data is predicted by using the updated machine learning model, thereby ensuring the accuracy of the predicted performance index of the network data.

An embodiment of the present invention further provides an information processing method applied to a second node, as shown in fig. 6, where the method includes:

step 601: reasoning the predicted value output by the trained machine learning model to obtain a reasoning result; evaluating the reasoning result to obtain an evaluation result;

step 602: judging whether to report the evaluation result to the first node or not; when the evaluation result is determined to be reported to the first node, generating first information; the first information represents relevant parameters of a machine learning model trained by the second node;

step 603: and sending the first information to the first node.

Here, in step 601, the second node infers the predicted value output by the trained machine learning model to obtain an inference result, evaluates the inference result to obtain an evaluation result, and then may generate first information based on the evaluation result and report the first information to the first node.

Here, in step 602, after the second node reports first information to the first node, the first node may determine, by using the first information, whether the machine learning model currently trained by the second node needs to be updated, so that the second node can update the currently trained machine learning model and predict index performance of network data by using the updated machine learning model, thereby ensuring accuracy of a prediction result.

In practical application, the second node may report the learning mode of the trained machine learning model to the first node, and thus, when the first node determines that the learning mode of the machine learning model of a certain second node needs to be updated, the first node may determine the model selection policy according to the learning mode reported by the second node.

Based on this, in an embodiment, the generating the first information includes:

In practical applications, when the first node determines that the learning mode of the machine learning model of the second node is updated, the second node may receive the first learning mode sent by the first node, and retrain the machine learning model by using the first learning mode.

Based on this, in an embodiment, the method further comprises:

receiving a first learning mode sent by the first node; the first learning mode is determined by the first node when the performance of the machine learning model trained by the second node is determined not to meet a preset condition;

retraining the machine learning model by using the first learning mode so as to enable the performance of the machine learning model to accord with a preset condition.

Here, the first learning mode includes at least one of:

a transfer learning mode,

An autonomous learning mode,

A combination of a migratory learning mode and an autonomous learning mode.

Here, the receiving, by the second node, the first Learning mode sent by the first node through a Learning mode instruction message specifically includes:

when the first Learning mode is the migration Learning mode, the content of the Learning mode instruction message may include the name or number of the DU-a node, the recommended Learning mode is the migration Learning mode, the migrated source machine Learning model, the name/number of the node corresponding to the source machine Learning model, the number of the source machine Learning model, and so on.

When the first Learning mode is the autonomous Learning mode, the content of the Learning mode instruction message may include the name or number of the DU-a node, and the recommended Learning mode is the autonomous Learning mode.

When the first Learning mode is a combination of a migration Learning mode and an autonomous Learning mode, the content of the Learning mode instruction message may include the name or number of the DU-a node, and the recommended Learning mode is a combination of the autonomous Learning mode and the migration Learning mode.

In practical application, the second node can update the learning mode of the machine learning model by itself, and can also update the learning mode of the machine learning model by the third node.

Based on this, in an embodiment, the method further comprises:

notifying a third node of the first learning mode;

Here, in practical application, the third node and the second node may be two separate physical entities or may be one physical entity in a set.

Here, the first learning mode may also be notified to a third node by the first node.

Fig. 7a and 7b are schematic diagrams of modules corresponding to a first node, a second node, and a third node, where as shown in fig. 7a, the first node is provided with a model evaluation module, the second node is provided with an online inference module, and the third node is provided with an AI training module.

And the online reasoning module of the second node is used for performing online reasoning on all the selected models currently based on the real-time samples, performing performance evaluation on the models according to the reasoning result and uploading the models and the evaluation result to the first node.

And the model evaluation module of the first node is used for judging which nodes have the standard model and which nodes do not have the standard model according to the model performance evaluation result and the set threshold. For the standard-reaching nodes, indicating and recommending the adopted models and learning modes (multiple types) to an AI training module, analyzing the similarity of the models among the nodes, and for the nodes with similar models, simultaneously matching a migration mode for partial nodes; for an substandard node, the new learning mode(s) for that node is recommended to the AI training module.

And the AI training module of the third node is used for generating a usable model set for each second node according to the learning mode suggestion indication of the first node, wherein the model can be an autonomous learning model only, can also be a plurality of migration learning models, can also be an autonomous model plus a migration learning model, and can also store the models in time sequence, and select a currently suitable model for use according to the indication of the first node.

As shown in fig. 7b, the AI training entity and the online reasoning entity of each DU are dispersed in different physical entities, the AI training module is a centralized AI server, and the model evaluation module is located in the CU.

In an example, taking the first node as a CU, the second node as a DU, and the third node as an AI training node as an example, an interaction process between the first node and the second node is described, as shown in fig. 8, including:

step 801: the DU-A infers a predicted value output by the trained machine learning model to obtain an inference result; and evaluating the reasoning result to obtain an evaluation result.

Here, DU-a is the second node.

Step 802: judging whether to report the evaluation result to the CU; when it is determined to report the evaluation result to the CU, step 803 is executed.

Step 803: generating first information; and sent to the CUs.

Here, the CU is the first node. DU-a may configure trigger conditions for uploading evaluation results, including periodicity, eventuality, e.g., model performance changes emutexceed a threshold, or model performance absolute values emutexceed a threshold; and when the trigger condition of uploading the evaluation result is met, reporting the first information to the CU.

Here, the first information may specifically be a Model evaluation results update message. The contents of the Model evaluation results update may include: node name, model number, model content, model learning mode (autonomous or migration), model evaluation result, model generation time; and if the model is the migration model, the node name/number of the migrated source machine learning model and the number of the migrated source machine learning model are also included.

Here, after receiving the first information, the CU may store the performance evaluation result of the corresponding DU-a included in the first information, and update the current all model evaluation result set; if the message content includes a model generation time, the old model may be deleted based on that time.

Step 804: receiving a first learning mode sent by a CU; notifying an AI training node of the first learning mode;

the first learning mode is used for the AI training node to retrain the machine learning model so as to enable the performance of the machine learning model to meet a preset condition.

Here, DU-a has the following functions as a functional entity:

1. the time intervals or the number of the currently effective models or the model learning modes or the failure time of the models can be configured, the machine learning models are screened according to the configuration conditions, and all the machine learning models which meet the conditions are inferred, namely, the predicted values output by the trained machine learning models are inferred to obtain the inference results.

2. An evaluation period may be configured, and the inference result is evaluated according to the configured evaluation period, for example, the machine learning model with the best inference result is evaluated.

3. And configuring an online inference Model according to a first learning mode sent by the CU, synchronizing the configured online inference Model information to the self AI training module through a Model update response message, and retraining the machine learning Model by the self AI training module. After receiving the Model update response message, the AI training module of DU-a may only keep the machine learning Model configured and used by the node and delete other machine learning models. The content of the Model update response message may include a node name, a Model number, and a Model generation time.

Here, the AI training node serves as a functional entity, and specifically has the following functions:

(1) default generation of models in autonomous learning mode

(2) Receiving and parsing the Learning mode instruction message, generating one or more models for the node according to the indicated Learning mode, including autonomous Learning only (generating one model), transfer Learning only (one or more transfer models may be generated), and autonomous + transfer Learning (generating multiple models).

(3) The updated models are synchronized into the corresponding online inference module through the following messages,

model Update request: the content includes node name, task number, model content, model learning mode (autonomous or migration), and if the model is a migration model, the content also includes migration model source node name/number, migration source model number, and also may include model generation time.

(4) After the on-line reasoning module receives the message, the relevant information of all the models is stored under the node name.

The following describes the interaction process between the first node and the second node in detail with reference to specific embodiments.

In the following, the AI training server corresponds to the third node, DU-a and DU-B correspond to the second node, and CU corresponds to the first node.

The AI training server stores data samples of network performance indexes KPI (such as switching success rate), and trains a prediction model capable of predicting the switching success rate according to the stored samples. The stored model set is shown in table 5.

TABLE 5

The AI training server updates the above model to DU-A and DU-B, and the format of the related message is emutexemplified as follows:

model update request message: { node name: DU _ A; and (4) task numbering: predicting the switching success rate; model 1: { model number: m _ A _ s _ 1; a learning mode: autonomous learning; generation time: 2020-10-22hh mm: ss; the contents of the model are as follows: … … }; model 2: { … … }

DU _ B performs online reasoning using the m _ B _ s _1 model, and the performance of the computational model is, for example, 92% of prediction accuracy, and the result is sent to the CU, and the message content is exemplified as follows:

updating the evaluation result of the model: { node name: DU _ B; and (4) task numbering: predicting the switching success rate;

model 1: { model number: m _ B _ s _ 1; a learning mode: autonomous learning; generation time: 2020-10-22hh mm: ss; the contents of the model are as follows: … …, evaluation results: 92% }

}

DU _ a performs online reasoning using m _ a _ s _1 and m _ a _ B _ s _1 models, the performance of the computational model is for example 93% and 91% respectively for prediction accuracy, the result is sent to CU, and the message content is exemplified as follows:

updating the evaluation result of the model: { node name: DU _ A; and (4) task numbering: predicting the switching success rate;

model 1: { model number: m _ A _ s _ 1; a learning mode: autonomous learning; generation time: 2020-10-22hh mm: ss; the contents of the model are as follows: … …, evaluation results: 95% };

model 2: { model number: m _ A _ B _ s _ 1; a learning mode: transfer learning; generation time: 2020-10-22hh mm: ss; the contents of the model are as follows: … …, evaluation results: 91% }

}

The CU evaluates according to an achievement threshold of a prediction accuracy of 90%, and considers that the model m _ B _ s _1 of DU _ B is achieved, and both models of DU _ a are achieved, and further selects the model m _ a _ B _ s _1 for DU _ a according to the policy "data storage minimum" to use, and informs DU of the following through a message, where the message content is as follows:

and (3) indicating an online model:

{ node name: DU _ B; and task numbering: predicting the switching success rate; recommendation model { model number: m _ B _ s _ 1; … … }

{ node name: DU _ A; and (4) task numbering: predicting the switching success rate; recommendation model { model number: m _ A _ B _ s _ 1; … … }

If the prediction accuracy reaching threshold configured by the CU is 93%, it is determined that DU _ B does not reach the standard model, the learning mode of B is changed to "autonomous + migratory learning" through a message, and the message content is exemplified as follows:

learning mode indication:

{ node name: DU _ B; and (4) task numbering: predicting the switching success rate; learning mode { autonomous learning, migratory learning: { source model: m _ a _ s _1, source model performance: 95% }

After receiving the learning mode instruction, the AI training server learns a new model m _ B _ a _ s _1 for DU _ B according to the migrated machine learning model as follows:

TABLE 6

By adopting the technical scheme of the embodiment of the invention, the first node can judge whether the second node needs to update the machine learning model or not by utilizing the first information reported by the second node, and sends the corresponding first learning mode to the second node when the second node is determined to need to update the machine learning mode, so that the second node can update the machine learning model by utilizing the first learning mode to predict the performance index of the network data by utilizing the updated machine learning model, thereby ensuring the accuracy of the predicted performance index of the network data.

In order to implement the information processing method according to the embodiment of the present invention, an information processing apparatus is further provided according to the embodiment of the present invention, and fig. 9 is a schematic structural diagram of the information processing apparatus according to the embodiment of the present invention; as shown in fig. 9, the apparatus includes:

a receiving unit 91, configured to receive first information sent by at least two second nodes respectively; the first information represents relevant parameters of a machine learning model trained by the second node;

the first processing unit 92 is configured to, for each second node, determine, based on the first information, whether performance of a machine learning model trained by the corresponding second node meets a preset condition; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and informing the corresponding second node of the first learning mode;

In an embodiment, the first processing unit 92 is specifically configured to: the first information comprises a second learning mode of the second node training machine learning model;

In an embodiment, the first processing unit 92 is specifically configured to:

generating second information; the second information comprises the name or the number of the corresponding second node, a recommended first learning mode, a migrated source machine learning model and the name or the number of the node corresponding to the migrated source machine learning model;

and sending the second information to the corresponding second node.

Furthermore, according to at least one embodiment of the present invention, the first processing unit 92 is specifically configured to:

the first information comprises a second learning mode of the second node training machine learning model;

when it is determined that the learning modes of the machine learning models trained by the other nodes are not the transfer learning mode, taking the autonomous learning mode as a first learning mode corresponding to the corresponding second node;

In an embodiment, the first processing unit 92 is specifically configured to: generating third information; the third information comprises the name or number of the corresponding second node, and a recommended first learning mode;

and sending the third information to the corresponding second node.

Further in accordance with at least one embodiment of the present invention, the first information includes a second learning mode in which the second node trains a machine learning model; the first processing unit 92 is specifically configured to:

In an embodiment, the first processing unit 92 is specifically configured to: generating fourth information; the fourth information comprises the name or number of the corresponding second node, a recommended first learning mode, a migrated source machine learning model, and the name or number of the node corresponding to the migrated source machine learning model;

and sending the fourth information to the corresponding second node.

In practical applications, the receiving unit 91 may be implemented by a communication interface in an information processing apparatus; the first processing unit 92 may be implemented by a processor in an information processing apparatus.

It should be noted that: in the information processing apparatus provided in the above embodiment, when performing information processing, only the division of each program module is exemplified, and in practical applications, the processing may be distributed to different program modules according to needs, that is, the internal structure of the apparatus may be divided into different program modules to complete all or part of the processing described above. In addition, the information processing apparatus and the information processing method provided by the above embodiments belong to the same concept, and specific implementation processes thereof are described in the method embodiments and are not described herein again.

In order to implement the information processing method according to the embodiment of the present invention, an information processing apparatus is further provided according to the embodiment of the present invention, and fig. 10 is a schematic structural diagram of the information processing apparatus according to the embodiment of the present invention; as shown in fig. 10, the apparatus includes:

the second processing unit 101 is configured to perform inference on a predicted value output by the trained machine learning model to obtain an inference result; evaluating the reasoning result to obtain an evaluation result;

a third processing unit 102, configured to determine whether to report the evaluation result to the first node; when the evaluation result is determined to be reported to the first node, generating first information; the first information represents relevant parameters of a machine learning model trained by the second node;

a sending unit 103, configured to send the first information to the first node.

Furthermore, according to at least one embodiment of the present invention, the third processing unit 102 is specifically configured to:

In one embodiment, the apparatus further comprises:

the training unit is used for receiving a first learning mode sent by the first node; the first learning mode is determined by the first node when the performance of the machine learning model trained by the second node is determined not to meet a preset condition; and retraining the machine learning model by using the first learning mode so as to enable the performance of the machine learning model to meet a preset condition.

In an embodiment, the sending unit 103 is further configured to:

notifying a third node of the first learning mode;

In practical applications, the second processing unit 101 and the third processing unit 102 may be implemented by processors in an information processing apparatus. The transmission unit 103 may be implemented by a processor in an information processing apparatus.

In order to implement the information processing method according to the embodiment of the present invention, an information processing system is further provided according to the embodiment of the present invention, and fig. 11 is a schematic structural diagram of the information processing system according to the embodiment of the present invention; as shown in fig. 10, the system includes:

the second node 111 is used for reasoning the predicted value output by the trained machine learning model to obtain a reasoning result;

The first node 112 is configured to receive first information sent by at least two second nodes, respectively; the first information represents relevant parameters of a machine learning model trained by the second node; for each second node, judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and notifying the corresponding second node of the first learning mode; wherein the first learning mode is used for the corresponding second node to retrain the machine learning model so that the performance of the machine learning model meets a preset condition.

Here, the process of the first node 112 and the second node 111 performing information processing has been described above, and is not described herein again.

An embodiment of the present invention further provides a first node device, as shown in fig. 12, including:

a first communication interface 121 capable of performing information interaction with other devices;

the first processor 122 is connected to the first communication interface 121, and is configured to execute a method provided by one or more technical solutions of the foregoing smart device side when running a computer program. And the computer program is stored on the first memory 123.

It should be noted that: the specific processing procedures of the first processor 122 and the first communication interface 121 are detailed in the method embodiment, and are not described herein again.

Of course, in practice, the various components in the control server 120 are coupled together by a bus system 124. It will be appreciated that the bus system 114 is used to enable communications among the components. The bus system 124 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, however, the various buses are labeled as bus system 124 in fig. 12.

The first memory 123 in the embodiment of the present application is used to store various types of data to support the operation of the controller 110. Examples of such data include: any computer program for operating on the first node device 120.

The method disclosed in the embodiment of the present application may be applied to the first processor 122, or implemented by the first processor 122. The first processor 122 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be implemented by integrated logic circuits of hardware or instructions in the form of software in the first processor 122. The first Processor 122 may be a general purpose Processor, a Digital data Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc. The first processor 122 may implement or perform the methods, steps and logic blocks disclosed in the embodiments of the present application. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the method disclosed in the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium located in the first memory 123, and the first processor 122 reads the information in the first memory 123, and completes the steps of the foregoing method in combination with its hardware.

An embodiment of the present invention further provides a second node device, as shown in fig. 13, including:

a second communication interface 131 capable of performing information interaction with other devices;

and a second processor 132, connected to the second communication interface 131, configured to execute the method provided by one or more technical solutions of the foregoing smart device side when running a computer program. And the computer program is stored on the second memory 133.

It should be noted that: the specific processing procedures of the second processor 132 and the second communication interface 131 are detailed in the method embodiment, and are not described herein again.

Of course, in practice, the various components of the second node apparatus 130 are coupled together by the bus system 134. It will be appreciated that the bus system 134 is used to enable communications among the components. The bus system 124 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, however, the various buses are identified in FIG. 13 as the bus system 134.

The second memory 133 in the embodiment of the present application is used to store various types of data to support the operation of the terminal 120. Examples of such data include: any computer program for operating on the second node device 130.

The method disclosed in the embodiment of the present application may be applied to the second processor 132, or implemented by the second processor 132. The second processor 132 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be implemented by integrated logic circuits of hardware or instructions in the form of software in the second processor 132. The second Processor 132 may be a general purpose Processor, a Digital data Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc. The second processor 122 may implement or perform the methods, steps and logic blocks disclosed in the embodiments of the present application. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the method disclosed in the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium located in the second memory 123, and the second processor 122 reads the information in the second memory 123, and in combination with the hardware thereof, performs the steps of the foregoing method.

In an exemplary embodiment, the first node Device 120 and the second node Device 130 may be implemented by one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), Complex Programmable Logic Devices (CPLDs), Field Programmable Gate Arrays (FPGAs), general purpose processors (gpus), controllers, Micro Controllers (MCUs), microprocessors (microprocessors), or other electronic components for performing the foregoing methods.

It is understood that the memories (the first memory 123 and the second memory 133) of the embodiments of the present application may be volatile memories or nonvolatile memories, and may include both volatile and nonvolatile memories. Among them, the nonvolatile Memory may be a Read Only Memory (ROM), a Programmable Read Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a magnetic random access Memory (FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Compact Disc Read-Only Memory (CD-ROM); the magnetic surface storage may be disk storage or tape storage. Volatile Memory can be Random Access Memory (RAM), which acts as external cache Memory. By way of illustration and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Enhanced Synchronous Dynamic Random Access Memory (Enhanced DRAM), Synchronous Dynamic Random Access Memory (SLDRAM), Direct Memory (DRmb Access), and Random Access Memory (DRAM). The memories described in the embodiments of the present application are intended to comprise, without being limited to, these and any other suitable types of memory.

In an exemplary embodiment, the present invention further provides a storage medium, specifically a computer-readable storage medium, for example, the storage medium includes a first memory 113 storing a computer program, and the computer program is executable by the first processor 122 of the first node device 120 to perform the steps of the aforementioned control server side method. The computer readable storage medium may be Memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface Memory, optical disk, or CD-ROM.

It should be noted that: "first," "second," and the like are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

In addition, the technical solutions described in the embodiments of the present invention may be arbitrarily combined without conflict.

The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention.

Claims

1. An information processing method applied to a first node, the method comprising:

2. The method of claim 1, wherein the first information comprises a second learning mode of the second node training a machine learning model; the determining a first learning mode corresponding to the respective second node comprises:

when the second learning mode of the machine learning model trained by the other nodes is determined to be the transfer learning mode, taking the other nodes as nodes to be processed to obtain a plurality of nodes to be processed;

3. The method of claim 2, wherein notifying the respective second node of the first learning mode comprises:

and sending the second information to the corresponding second node.

4. The method of claim 1, wherein the first information comprises a second learning mode of the second node training a machine learning model; the determining a first learning mode corresponding to the respective second node comprises:

5. The method of claim 4, wherein notifying the respective second node of the first learning mode comprises:

and sending the third information to the corresponding second node.

6. The method of claim 1, wherein the first information comprises a second learning mode of the second node training a machine learning model; the determining a first learning mode corresponding to the respective second node comprises:

7. The method of claim 6, wherein the sending to the respective second node comprises:

and sending the fourth information to the corresponding second node.

8. An information processing method applied to a second node, the method comprising:

evaluating the reasoning result to obtain an evaluation result;

judging whether to report the evaluation result to the first node or not;

when the evaluation result is determined to be reported to the first node, generating first information; the first information represents relevant parameters of a machine learning model trained by the second node;

and sending the first information to the first node.

9. The method of claim 8, wherein generating the first information comprises:

10. The method of claim 8, further comprising:

receiving a first learning mode sent by the first node; the first learning mode is determined by the first node when the performance of the machine learning model trained by the second node is determined to be not in accordance with a preset condition;

and retraining the machine learning model by using the first learning mode so as to enable the performance of the machine learning model to meet a preset condition.

11. The method of claim 10, further comprising:

notifying a third node of the first learning mode;

12. An information processing apparatus characterized by comprising:

13. An information processing apparatus characterized by comprising:

a sending unit, configured to send the first information to the first node.

14. A first node device, comprising:

the first processor is used for judging whether the performance of the machine learning model trained by the corresponding second node meets a preset condition or not based on the first information for each second node; when the performance of the machine learning model trained by the corresponding second node is determined not to meet the preset condition, determining a first learning mode corresponding to the corresponding second node, and informing the corresponding second node of the first learning mode;

15. A second node device, comprising:

16. A first node apparatus comprising a processor and a memory for storing a computer program operable on the processor,

wherein the processor is adapted to perform the steps of the method of any one of claims 1 to 7 when running the computer program.

17. A second node device comprising a processor and a memory for storing a computer program capable of running on the processor,

wherein the processor is adapted to perform the steps of the method of any one of claims 8 to 11 when running the computer program.

18. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7 or carries out the steps of the method of any one of claims 8 to 11.