Method, device and machine readable medium related to machine learning in communication network

Abstract

The invention relates to a method, a device and a machine readable medium related to machine learning in a communication network. There is provided a method performed by a coordinating network entity in a communication network, comprising: sending a first request message from the coordinating network entity to a network registration entity in the communication network for requesting identification information of a plurality of candidate network entities capable of performing collaborative learning in the communication network; receiving, at the coordinating network entity, identification information of the plurality of candidate network entities from the network registration entity; sending a second request message comprising at least one query of additional information for the plurality of candidate network entities; selecting one or more network entities from the plurality of candidate network entities based on the one or more responses to the at least one query; and initiating, at one or more network entities selected from the plurality of candidate network entities, training the model using a machine learning algorithm as part of a collaborative learning process.

Description

[0001] This application is a divisional application of the PCT international application PCT / EP2020 / 072123, filed on August 6, 2020, with the invention name “Methods, devices and machine-readable media related to machine learning in communication networks”, which entered the Chinese national phase patent application 202080057773.7. Technical Field

[0002] Embodiments of the present disclosure relate to machine learning, and in particular to methods, apparatus, and machine-readable media related to machine learning in communication networks. Background Art

[0003] In a typical wireless communication network, wireless devices are connected to the core network through a radio access network. In the fifth generation (5G) wireless communication network, the core network operates according to a service-based architecture (SBA), in which services are provided by network functions via defined application interfaces (APIs). Network functions in the core network use a common protocol framework based on Hypertext Transfer Protocol 2 (HTTP / 2). In addition to providing services, network functions can also call services in other network functions through these APIs. Examples of core network functions in the 5G architecture include access and mobility management function (AMF), authentication server function (AUSF), session management function (SMF), policy and charging function (PCF), unified data management (UDM), and operation, administration and maintenance (OAM). For example, AMF can request subscriber authentication data from AUSF for this purpose by calling functions in AUSF's API.

[0004] Efforts are underway to automate 5G networks, aiming to provide zero-touch, fully automated wireless communication networks (i.e., networks that require as little human intervention as possible during operation). One way to achieve this is to use the large amounts of data collected in wireless communication networks in conjunction with machine learning algorithms to develop models for providing network services.

[0005] A Network Data Analysis (NWDA) framework has been established to define mechanisms and associated functions for data collection in 5G networks. Further enhancements to the framework are described in 3GPP document TS 23.288 v 16.0.0. The NWDA framework is centered around a Network Data Analysis Function (NWDAF) that collects data from other network functions in the network. The NWDAF also provides services to service consumers (e.g. other network functions). For example, services include retrieving data or making predictions based on the data collated at the NWDAF.

[0006] Figure 1The NWDAF 102 is shown connected to a network function (NF) 104. As shown, the network function 104 can be any suitable network function (e.g., AMF, AUSF, or any other network function). In order to collect data from the network function 104, the NWDAF 102 is connected to an event exposure function at the network function via an Nnf reference point (as detailed in 3GPP documents TS 23.502 v16.0.2 and TS 23.288 v 16.0.0). The NWDAF 102 can then receive data from the network function via the Nnf reference point by subscribing to reports from the network function or by requesting data from the network function. The timing of any report can be determined by a timeout (e.g., timer expiration) or can be triggered by an event (e.g., receipt of a request). The type of data that the NWDAF 102 can request from the network function can be standardized.

[0007] In order for the network function 104 to be discoverable by the NWDAF 102 (or any other service consumer, such as another network function), the network function 104 is registered with a Network Function Repository Function (NRF).

[0008] Figure 2 A diagram of an NRF 208 connected to three network functions NF A 202, NF B 204, and NF C 206 registered at the NRF 208 is shown. The NRF 208 may be pre-configured with information about the network functions 202-206, or each of the network functions 202-206 may have performed a network registration procedure with the NRF 208 to register at the NRF 208. Once a network function is registered at the NRF 208, another entity in the network may discover the network function by invoking a discovery function at the NRF 208. Thus, for example, NF B 204 may discover NF A 202 and NF C 206 by invoking a discovery function at the NRF 208.

[0009] As mentioned above, data collection has the potential to be a powerful tool for 5G networks when combined with machine learning. Machine learning in 5G network contexts is typically large-scale and can be performed in cloud (virtualized) environments that prioritize performance and security. In practice, this means that the data available for training models using machine learning may be distributed across many entities in the network, and ideally, this data should be collated at one network entity for use in developing models using machine learning. Collation of these datasets at a single network entity can be slow and resource intensive, which is problematic for time-critical applications. Additionally, some applications require the use of datasets that include sensitive or private data, and collation of this data at a single network entity can have security implications. Summary of the invention

[0010] Embodiments of the present disclosure address these and other issues.

[0011] In one aspect, a method performed by a coordination network entity in a communication network is provided. The method includes sending a request message for identification information of multiple candidate network entities in the communication network that can perform collaborative learning to a network registration entity in the communication network. The method also includes receiving identification information of the multiple candidate network entities from the network registration entity; and initiating training of a model using a machine learning algorithm at one or more network entities in the multiple candidate network entities as part of a collaborative learning process.

[0012] In another aspect, the present disclosure provides a coordination network entity to perform the above method. In another aspect, the present disclosure provides a computer program for performing the above method. A computer program product including the computer program is also provided.

[0013] In another aspect, a method performed by a network registration entity in a communication network is provided. The method includes receiving a request message from a coordination network entity in the communication network, the request message requesting identification information of multiple candidate network entities in the communication network that can perform collaborative learning. The method also includes identifying two or more candidate network entities that can perform collaborative learning from multiple network entities registered at the network registration entity; and sending an indication of the two or more candidate network entities to the coordination network entity.

[0014] In another aspect, the present disclosure provides a network registration entity to perform the above method. In another aspect, the present disclosure provides a computer program for performing the above method. A computer program product including the computer program is also provided.

[0015] On the other hand, a coordination network entity is provided. The coordination network entity includes a processing circuit and a machine-readable medium storing instructions, and when the instructions are executed by the processing circuit, the coordination network entity sends a request message for identification information of multiple candidate network entities in the communication network that can perform collaborative learning to a network registration entity in the communication network. The coordination network entity also receives identification information of the multiple candidate network entities from the network registration entity, and initiates training of a model using a machine learning algorithm at one or more of the multiple candidate network entities as part of the collaborative learning process.

[0016] On the other hand, a network registration entity is provided. The network registration entity includes a processing circuit and a machine-readable medium storing instructions, and the instructions, when executed by the processing circuit, cause the network registration entity to receive a request message from a coordination network entity in a communication network. The request message requests identification information of multiple candidate network entities in the communication network that can perform collaborative learning. The network registration entity is also caused to identify one or more candidate network entities that can perform collaborative learning from multiple network entities registered at the network registration entity 900, and send an indication of the one or more candidate network entities to the coordination network entity. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] For a better understanding of the examples of the present disclosure and for more clearly showing how the examples may be effectively implemented, reference will now be made, by way of example only, to the accompanying drawings, in which:

[0018] Figure 1 showing a network data analysis function connected to a network function;

[0019] Figure 2 A network function repository function connected to three network functions is shown;

[0020] Figure 3 A system according to an embodiment of the present disclosure is shown;

[0021] Figure 4 is a schematic signaling diagram according to an embodiment of the present disclosure;

[0022] Figure 5 is a flow chart of a method according to an embodiment of the present disclosure;

[0023] Figure 6 is a flow chart of a method according to an embodiment of the present disclosure;

[0024] Figure 7 and Figure 8 is a schematic diagram of a coordination network entity according to an embodiment of the present disclosure; and

[0025] Fig. 9 and Fig.10 is a schematic diagram of a network registration entity according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0026] Figure 3A system 300 in a communication network according to an embodiment of the present disclosure is shown. For example, one or more entities of the system may form part of a core network in the communication network. The core network may be a fifth generation (5G) core network (5G CN). The communication network may implement any suitable communication protocol or technology, such as Global System for Mobile Communications (GSM), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), New Radio (NR), WiFi, WiMAX, or Bluetooth wireless technology. In a specific example, the network forms part of a cellular telecommunications network, such as the type developed by the Third Generation Partnership Project (3GPP). Those skilled in the art will appreciate that the system 300 may include, for clarity, Figure 3 Other components omitted.

[0027] The system 300 includes at least two network entities or network functions (NFs). In the illustrated embodiment, three network entities NF A 302, NF B 304, and NF C 306 are shown, but those skilled in the art will appreciate that the system 300 may include more network entities than shown. The network entities 302-306 are configured to provide one or more services. The network entity may be any type or combination of types of network entities or network functions. For example, one or more of the network entities 302-306 may include core network entities or functions, such as access and mobility management functions (AMFs), authentication server functions (AUSFs), session management functions (SMFs), policy control functions (PCFs), and / or unified data management (UDM) functions. Alternatively or additionally, one or more of the network entities 302-306 may be implemented in an entity outside the core network, such as a radio access network node (e.g., a base station such as a gNB, eNB, etc., or a portion thereof, such as a central unit or a distributed unit). The network entities 302-306 may be implemented in hardware, software, or a combination of hardware and software.

[0028] Each of the network entities 302-306 is registered at a network registration entity 310, which also forms part of the system 300. In this example, the network registration entity is a network function repository function (NRF) 310. However, those skilled in the art will appreciate that the network registration entity may be any suitable network entity that provides registration and discovery for network entity services. The NRF 310 may therefore store information for each network entity 302-306 registered therein. The stored information may include one or more of the following: the type of each of the network entities 302-306; the network address (e.g., IP address) of the network entity; the services provided by the network entity; and the capabilities of the network entity. Therefore, once registered at the NRF 310, the network entities 302-306 may be discovered by other entities in the network.

[0029] The system 300 also includes a network data and analysis function (NWDAF) 308. The NWDAF 308 is configured to collect network data from one or more network entities and provide network data analysis information to network entities that request or subscribe to receive network data analysis information. For example, the NWDAF can provide information related to network services or usage (e.g., predicted load information or statistics related to historical load information). The network data analysis information provided by the NWDAF can be, for example, specific to the entire network, or specific to a portion of the network, such as a network entity or a network slice. In this context, a network slice can be a logical partition (e.g., a virtual network) in a communication network. A network slice can be dedicated to a specific use case or end user. For example, a network slice can include one or more network functions for a specific use case or end user. In a specific example, a network slice can be defined in hardware. For example, a network slice can include a group of servers dedicated to a specific use case or end user. A network slice can be isolated from or independent of other parts of a communication network. For example, the transmission of information between a network slice and other parts of the communication network (e.g., other network slices) can be prohibited or restricted.

[0030] The network data analysis information provided by the NWDAF 308 may include forecast data (e.g., an indication of a forecasted load of a network function) and / or historical data (e.g., an average number of wireless devices in a cell in the communication network). The network data analysis information provided by the NWDAF may include, for example, performance information (e.g., a ratio of successful handovers to failed handovers, a ratio of successful setups to failed setups of protocol data unit (PDU) sessions, a number of wireless devices in an area, an indication of resource usage, etc.).

[0031] As described above, communication networks are becoming increasingly automated, and network designers seek to minimize the level of human intervention required during operation. One way to achieve this is to use data collected in the communication network to train models using machine learning, and use these models in the control of the communication network. As the communication network continues to obtain data during operation, the models can be updated and adjusted to suit the needs of the network. However, as described above, conventional methods for implementing machine learning in communication networks require that the data used to train the models be collated at one network entity. Collation of this data at a single network entity (e.g., NWDAF 308) can be slow and resource intensive, and can be problematic if the data is sensitive in nature.

[0032] Aspects of the present disclosure address these and other issues.

[0033] In one aspect, a collaborative (e.g., federated) learning process is used to train a model using machine learning. Instead of collating training data for training a model at a single network entity, instances of the model are trained locally at multiple network functions to obtain local updates to model parameters at each network entity. The local model updates are collated and combined at an aggregator network entity (e.g., NWDAF) to obtain a combined model update. In this way, data from multiple entities in a communication network is used to train a model using machine learning while minimizing resource overhead and reducing security risks.

[0034] Therefore, in Figure 3 In the illustrated system 300, the NWDAF 308 initiates the training of the model using machine learning at each network function NF A 302, NF B 304, and NF C 306. For example, the NWDAF 308 may send a message to each of the network functions 302-306, the message instructing the network function to train the model using machine learning. The message may include a copy of the model (e.g., a global copy that is common to each of the network functions 302-306), or each of the network functions 302-306 may be pre-configured with a copy of the model. In the latter case, the message may include an indicator of which model is to be trained. The message may specify the type of machine learning algorithm to be used by the network entity. Alternatively, the network entity 302-306 may be pre-configured with the type of machine learning algorithm to be used for the model.

[0035] Upon receiving the message from NWDAF 308, each network entity 302-306 trains the model by inputting the training data into the machine learning algorithm to obtain local model updates to the values ​​of one or more parameters of the model. The training data may be data unique to the network entity. For example, the training data may include data obtained from measurements performed by the network function and / or data collected by the network function from other network entities (e.g., data obtained from measurements performed by one or more other network entities).

[0036] Each of the network entities 302-306 sends a local model update to the NWDAF 308. The local model update may include an updated value of a model parameter, or the local model update may include an indication of a change in the value of a model parameter, such as a difference between a previous value of the parameter and an updated value of the parameter.

[0037] Transmissions between network entities 302-306 and NWDAF 308 may be direct (eg, NWDAF 308 sends directly to the network entity) or may occur via an intermediate network entity. For example, transmissions between network functions 302-306 and NWDAF 308 may occur via operations, administration, and maintenance functions (OAM) 312.

[0038] The NWDAF 308 thus receives local model updates from each of the network entities 302-306. The NWDAF 308 combines the model updates received from the network entities 302-306 to obtain a combined model update. The NWDAF 308 may combine the model updates using any suitable operation. For example, the NWDAF 308 may average the received local model updates to obtain an average model update. In another example, the average may be a weighted average, where updates from different network entities are assigned different weights.

[0039] The NWDAF 308 sends the combined model update to one or more network entities in the network. For example, the NWDAF 308 may send the combined model update to each of the network entities 302-306. In a specific example, in addition to the network entities 302-306 used to train the model, the combined model update may be sent to one or more additional network entities.

[0040] This process may be repeated one or more times. For example, the process may be repeated until the local model updates received from each of the network entities 302-306 are consistent with each other within a predetermined tolerance. In another example, the process may be repeated until the combined model updates converge (i.e., the combined model updates are consistent with the previous combined model updates) to within a predetermined tolerance.

[0041] Therefore, collaborative (e.g., federated) learning can be applied in communication networks (especially core networks in communication networks) to reduce latency, minimize resource overhead, and reduce the risk of security issues.

[0042] As described above, as part of applying collaborative learning to a communication network, a network entity initiates training of a machine learning model at a plurality of other network entities in the communication network. However, a communication network typically includes a large number of network entities, and only a small portion of these network entities may be configured to support collaborative learning. Furthermore, participation in the training of a particular model using a particular machine learning algorithm may be associated with specific hardware or software requirements that are met by only a subset of the network entities.

[0043] Embodiments of the present disclosure provide a method, apparatus, and machine-readable medium for selecting a network entity to perform collaborative learning. Specifically, a coordination network entity in a communication network sends a request message for identification information of multiple candidate network entities in the communication network that can perform collaborative learning to a network registration entity in the communication network. Then, the coordination entity receives identification information of multiple candidate network entities from the network registration entity, and initiates training of a model using a machine learning algorithm at one or more of the multiple candidate network entities as part of a collaborative learning process.

[0044] Therefore, the embodiments described herein provide an effective method for selecting network entities to perform a collaborative learning process. Specifically, by identifying candidate network entities that are capable of performing collaborative learning, the embodiments described herein prevent the selection of network entities that are not capable of performing collaborative learning to perform the collaborative learning process, thereby minimizing the risk of errors when model updates obtained during the collaborative learning process are aggregated and combined. Therefore, embodiments of the present disclosure provide a more reliable method for selecting network entities to perform collaborative learning in a communication network.

[0045] Figure 4 is a schematic signaling diagram according to an embodiment of the present disclosure. The signaling involves a coordination network entity, a network registration entity, and a network entity or function. The coordination network entity may include an NWDAF, such as the one described above with respect to Figure 3 The network registration entity may include an NRF, such as the NWDAF 308 described above. Figure 3 In the illustrated embodiment, signaling also involves OAM, such as the above Figure 3 Described is OAM 312. However, other embodiments may not involve OAM. Figure 4 The signaling shown in allows the coordinating network entity (hereinafter referred to as NWDAF 308) to select one or more network entities to participate in a collaborative learning process such as federated learning. Thus, it can be performed when the NWDAF 308 decides to develop or further train a model. Figure 4 Signaling in .

[0046] exist Figure 4 Prior to the signaling shown, one or more network entities (including NF A 302 shown in the figure) register with NRF 310. When registering with NRF 310, the network entity provides information related to the services provided by the network entity (e.g., the type of network entity or the functions performed thereby) and / or the capabilities of the network entity. Such information can be stored by NRF 310 and associated with an identifier (e.g., a unique number within the network) that allows identification and addressing of the network entity. Further information about the stored information and identifiers is described below.

[0047] The process begins with NWDAF 308 sending a first request message 400 for identification information of a plurality of candidate network entities in the communication network that are capable of performing collaborative learning to NRF 310. The first request message 400 may be sent directly from NWDAF 308 to NRF 310 (as shown) or indirectly via one or more intermediate entities in the communication network. The first request message 400 thus includes a request for NRF 310 to provide a list of network entities that are capable of performing collaborative learning (e.g., federated learning).

[0048] The first request message 400 may include one or more first selection criteria for selecting a candidate network entity to perform the collaborative learning process. The one or more first selection criteria may include at least one primary criterion related to the capability of the candidate network entity to perform the collaborative learning process.

[0049] The at least one primary criterion may specify one or more specific types of network entities. For example, the first request message 400 may indicate that each candidate network entity must be an access and mobility management function (AMF), or a dedicated machine learning network function. The at least one primary criterion may additionally or alternatively specify services available at or provided by each candidate network entity. That is, the first request message 400 may specify that each network entity selected as a candidate network entity is capable of providing a specific service. For example, the first request message 400 may specify that each network entity selected as a candidate network entity is capable of providing subscriber authentication data or a specific type of collaborative learning.

[0050] At least one primary criterion may relate to hardware and / or software requirements of the candidate network entities. For example, at least one primary criterion may specify that each candidate network entity must have a graphics processing unit (GPU) or other suitable hardware that can be used to train a model using machine learning.

[0051] Additionally or alternatively, the one or more first selection criteria in the first request message 400 may include at least one secondary criterion related to the ability of the candidate network entity to support the query type. That is, the at least one secondary criterion is related to whether the candidate network entity is configured to respond to a specific type of query. Examples of different types of queries are discussed in more detail below, but may include, for example, queries from a specific type of network entity (e.g., NWDAF 308 or OAM 312), and / or queries for specific types of information.

[0052] The NRF 310 identifies two or more candidate network entities capable of performing collaborative learning from a plurality of network entities registered at a network registration entity. The two or more candidate network entities may be, for example, in a core network of the communication network. Alternatively, at least one of the two or more candidate network entities may be, for example, in another part of the communication network (e.g., a radio access network, a backhaul network, etc.).

[0053] The NRF 310 may store a profile for each network entity registered with it. Each profile may include information related to the services provided by the network entity (e.g., the type of network entity or the functions performed thereby) and / or the capabilities of the network entity. Such information may be stored by the NRF 310 and associated with an identifier (e.g., a unique number within the network) or some other means that allows identification and addressing of the network entity. The profile may include an indication of one or more services that the network entity is capable of providing. For example, the profile may indicate the type of network entity, e.g., whether the network entity is a PCF, ASF, or a dedicated machine learning network function or entity.

[0054] The capability information may include an indication of whether the network entity is capable of performing collaborative (eg, federated) learning. Thus, the NRF 310 may use profiles of multiple network entities registered at a network registration entity to identify a network entity capable of performing collaborative learning.

[0055] The capability information included in the profile of the network entity registered at the NRF 310 may also include an indication of whether the network entity is configured to respond to a query type. For example, the profile of the network entity may specify that the network entity is capable of responding to queries using a protocol and / or format used by a particular type of network entity (e.g., from the NWDAF 308 or OAM 312). Additionally or alternatively, the profile of the network entity may specify that the network entity is capable of responding to queries related to storage available at the network entity.

[0056] As described above, the first request message 400 may include one or more first selection criteria for selecting a candidate network entity from two or more network entities to train a model using machine learning as part of a collaborative learning process. Thus, as part of identifying two or more candidate network entities capable of performing collaborative learning, the NRF 310 may identify two or more candidate network entities that meet the one or more first selection criteria by searching the profiles of the network entities registered therewith.

[0057] NRF 310 sends an indication of two or more candidate network entities to NWDAF 308 in a first response message 402. For example, the indication may include identification information of each of the two or more candidate network entities. First response message 402 may also include addressing information of the two or more candidate network entities (e.g., information that enables the two or more candidate network entities to be addressed in the communication network). The identification information may correspond to the addressing information. For example, an identifier stored in a profile may correspond to the identification information and / or the addressing information of the network entity.

[0058] In some embodiments, after receiving an indication of two or more candidate network entities from NRF 310, NWDAF 308 initiates training of a model using a machine learning algorithm at the two or more candidate network entities as part of a collaborative learning process. For example, NWDAF 308 may send a trigger message to each candidate network entity, the trigger message instructing the candidate network entity to train the model using a machine learning algorithm. For example, the trigger message may include information related to the model (e.g., model parameters and / or identification information of the model) and / or an indication of the machine learning algorithm to be used. Alternatively, the candidate network entity may be preconfigured with some or all of this information.

[0059] However, in other embodiments, NWDAF 308 sends a second request message 404 that includes at least one query for additional information for each of the two or more candidate network entities. The at least one query may include one or more second selection criteria for selecting a candidate network entity.

[0060] In embodiments where first request message 400 includes at least one secondary criterion related to the candidate network entity's ability to support a query type, second request message 404 may include a query of the type specified in first request message 400. Thus, for example, if first request message 400 includes a criterion specifying that the candidate network entity must be able to respond to queries related to hardware at the candidate network entity, second request message 400 may specify that the candidate network entity has a particular type of hardware (e.g., a graphics processing unit).

[0061] Those skilled in the art will appreciate that the at least one query may take any suitable form. For example, the at least one query may include event filter information such as described in TS 23.502 v 16.1.1. Other suitable formats for queries may be found in, for example, TS 28.550 v 16.1.0, but those skilled in the art will appreciate that the present disclosure is not limited thereto.

[0062] At least one query may be related to the configuration of the candidate network entity, for example, what hardware and / or software is present at the candidate network entity or what hardware and / or software is used by the candidate network entity. For example, at least one query may request information related to the number of graphics processing units (GPUs) at the candidate network entity. In another example, at least one query may include one or more second selection criteria that specify a minimum number of graphics processing units (GPUs) at the candidate network entity. In another example, at least one query may include one or more selection criteria that specify a specific software version to be supported by the candidate network entity. Examples of queries related to the configuration of the candidate network entity include queries related to one or more of: software version, licensing, neighbor relations, one or more configuration parameters, and the type and / or quantity of hardware and / or software at the candidate network entity.

[0063] In certain embodiments, at least one query related to the configuration of the candidate network entity may be related to computing resources available at the candidate network entity. In this context, computing resources may include, for example, memory available at the network entity (e.g., random access memory or storage device), processing power at the network entity, and / or any other suitable computing resources. For example, at least one query may include a second selection criterion that specifies that the candidate network entity has a minimum amount of available memory. In another example, at least one query may include a second selection criterion that specifies that the candidate network entity has a minimum amount of available processing power.

[0064] At least one query for the candidate network entity may additionally or alternatively be related to performance requirements of the candidate network entity. Queries related to performance requirements may include, for example, queries related to performance measurement (PM) processes, key performance indicators (KPIs), and / or any other suitable performance metrics of the candidate network entity. One or more second selection criteria may therefore be related to thresholds for specific performance measurements, key performance indicators, and / or other performance metrics.

[0065] The performance requirements may be related to the historical or predicted performance of the candidate network entity. The performance requirements may be related to, for example, the use of a service provided by the candidate network entity (e.g., the number of users of the service, the number of times the service has been used), the traffic flow of the network entity, and / or the resources used by the candidate network entity (e.g., memory, processing power, and / or network resources). The performance requirements may be specific to a particular time period. For example, a performance requirement may specify a minimum number of predicted users within the next hour. In another example, a performance requirement may specify a threshold number of average users during the 12:00-12:15 time period last week.

[0066] At least one query to a candidate network entity may be related to the availability of training data at the candidate network entity. In order to train a model using a machine learning algorithm, the network entity inputs data (training data) into the machine learning algorithm. The training data of the candidate network entity may, for example, include data obtained from measurements performed by the candidate network entity and / or data collected by the candidate network entity from other network entities (e.g., data obtained from measurements performed by one or more other network entities). A candidate network entity that supports queries related to the availability of training data may be configured to indicate whether suitable training data is available at the candidate network entity in response to receiving a query of this type. For example, the query may be related to the number, age (i.e., an indication of when the data was measured or obtained), quantity, or content of the training data at the candidate network entity. Thus, for example, a candidate network entity may receive a query for information about the number of training data at the candidate network entity, and the candidate network entity may respond with an indication of the number of data samples available at the candidate network entity.

[0067] At least one query to a candidate network entity may be related to the performance of the candidate network entity when training a model using machine learning. Therefore, the candidate network entity may be configured to obtain one or more metrics indicating the performance of a preliminary model developed using a machine learning algorithm in response to receiving a query of this type. That is, the candidate network entity may be configured to test the performance of the model developed at the candidate network entity using a validation or test data set. The validation or test data set may be a subset of the training data available at the candidate network entity. Alternatively, the validation or test data set may be provided to the candidate network entity (e.g., by NWDAF308). The metric may include an indicator of the accuracy and / or precision of the model. Additionally or alternatively, the metric may indicate the efficiency of training the model at the candidate network entity. For example, the metric may include an indicator of the resources (memory, processing power, etc.) used to train the preliminary model at the candidate network entity and / or the time (e.g., processor time, wall clock time, etc.) used to train the preliminary model.

[0068] NWDAF 308 sends a second request message 404 including at least one query for additional information to OAM 312. NWDAF 308 may therefore send a single second request message 404 to OAM 312 for all candidate network entities. Alternatively, NWDAF 308 may send two or more second request messages to OAM 312 (e.g., one second request message for each candidate network entity).

[0069] OAM 312 receives a second request message 404 including at least one query from NWDAF 308. In some embodiments, OAM 312 stores information required to answer at least one query for each candidate network device (e.g., in a cache). For example, OAM 312 may store network traffic load information for each candidate network device. Thus, if at least one query specifies that a candidate device must have a network traffic load of less than 50%, OAM 312 may determine which candidate network devices meet the requirement and send an indication to NWDAF 308 of which candidate network devices meet the requirement.

[0070] Alternatively, if OAM 312 does not store information required to respond to at least one query for all candidate network devices, OAM 312 may generate a third request message 406 based on the at least one query and send the third request message 406 to at least one of the two or more candidate network entities. The third request message 406 may include at least one query, or the third request message 406 may be based at least in part on the second request message 404. For example, OAM 312 may store information required to respond to a portion of at least one query, but lack information required to respond to the remainder of the query. For example, the at least one query may be related to an average network traffic load and an average number of users of the candidate network entity, and OAM 312 may store the average network traffic load of the candidate network entity, but not the average number of users. Therefore, the third request message 406 may include a request for the remaining information required to answer the query. In an alternative embodiment, OAM 312 forwards the second request message 404 to each candidate network entity as a third request message 406.

[0071] Each candidate network entity that receives the third request message 406 from OAM 312 may send a second response message 408 to OAM 312, the message 408 including the additional information requested in the at least one query. For example, the third request message 406 may include a query related to the version of the software type supported by the candidate network entity. The candidate network entity may therefore indicate in the response message 408 the software version it is configured to use.

[0072] In an embodiment where at least one query includes one or more second selection criteria, each candidate network entity that receives the third request message 406 may indicate in the second response message 408 whether it meets the one or more second selection criteria included in the at least one query. For example, the candidate network entity may use one or more Boolean parameters to indicate which of the one or more second selection criteria is met (e.g., the second response message 408 may include a Boolean parameter for each second selection criterion). In one example, a candidate network entity that meets all of the one or more second selection criteria returns a true value (e.g., a logical value of 1). In another example, a candidate network entity that does not meet at least one of the one or more second selection criteria returns a false value (e.g., a logical value of 0). For example, at least one query may include a second selection criterion that the candidate network entity has collected at least 1,000 data samples (e.g., for training a model). Therefore, a candidate network entity that has only collected 800 data samples returns a false value in the second response message 408.

[0073] In certain embodiments, if at least one query includes one or more second selection criteria, the candidate network entity may respond to the third request message 406 only if the candidate network entity satisfies each of the one or more second selection criteria. Thus, in the above example, a candidate network entity that has collected less than 1000 data samples may not respond to the third request message 406. OAM 312 may therefore treat the lack of a response from the candidate network entity as an indication that the candidate network entity does not satisfy the one or more second selection criteria.

[0074] OAM 312 thus receives a second response message 408 from the one or more candidate network entities. OAM 312 sends a third response message 410 to NWDAF 308 based on the second response message 408. For example, OAM 312 may forward the second response message 408 to NWDAF 308 in the third response message 410. Alternatively, OAM 312 may extract the additional information requested in the at least one query from the second response message 408 and send the additional information to NWDAF 308 in the third response message 410.

[0075] If the at least one query includes one or more second selection criteria, OAM 312 may send a third response message 410 to NWDAF 308, wherein the third response message 410 includes an indication of which of the two or more candidate network entities satisfy the one or more second selection criteria. Thus, for example, the third response message 410 may include identification information of only the candidate network entities that satisfy the one or more second selection criteria. Alternatively, the third response message may include identification information of all candidate network entities initially identified to OAM 312 (in request message 404) and an indication (e.g., a Boolean value) of whether each candidate network entity satisfies the one or more second selection criteria.

[0076] In certain embodiments, OAM 312 may send indications for only a subset of candidate network entities that meet one or more second selection criteria. For example, OAM 312 may be preconfigured with a maximum number of candidate network entities to report to NWDAF 308. Alternatively, NWDAF 308 may indicate in the second request message 404 the maximum number of candidate network entities to report.

[0077] Thus, only those subsets of network entities that meet the second selection criteria may be reported to NWDAF 308. OAM 312 may select the subset of candidate network entities using any suitable method. For example, the subset of candidate network entities may be randomly selected. Alternatively, OAM 312 may select the subset of candidate network entities based on characteristics of the candidate network entities. Information related to the characteristics of the candidate network entities may be stored at OAM 312 (e.g., in a cache), or OAM may, for example, obtain the information from the candidate network entities (e.g., in a second response message 408).

[0078] OAM 312 may rank the candidate network entities according to certain characteristics and select the best (e.g., highest) ranked candidate network entity. A secure multi-party computation algorithm (e.g., Yao's millionaire's problem algorithm) may be used to compare the characteristics of multiple candidate network entities while minimizing the information shared between the network entities. For example, OAM 312 may select a subset of candidate network entities with the largest training data set. Thus, a secure multi-party computation algorithm may be used to determine which candidate network entities have the largest training data set without revealing the size of the training data set to OAM 312.

[0079] In other examples, OAM 312 may select a subset of candidate network entities based on the location of the candidate network entities in the network. For example, OAM 312 may prioritize candidate network entities in the same network slice.

[0080] Thus, in certain embodiments, OAM 312 selects a subset of candidate network entities based on one or more second response messages 408 received from the candidate network entities. OAM 312 then sends a third response message 410 to NWDAF. Using the third response message 410, NWDAF 308 selects one or more candidate network entities to participate in a collaborative learning process to train a model using a machine learning algorithm.

[0081] In some embodiments, the third response message 410 includes the additional information requested in the at least one query. The NWDAF 308 can therefore select one or more candidate network entities based on the additional information. For example, if the additional information includes an indication of the number of data samples available at each candidate network entity for training a model using machine learning, the NWDAF 308 can select one or more candidate network entities with the largest number of data samples. In another example, the NWDAF 308 can apply one or more second selection criteria (e.g., the one or more second selection criteria described above) to select one or more candidate network entities. In this example, the NWDAF 308, rather than the candidate network entity, determines whether the candidate network entity satisfies the one or more second selection criteria.

[0082] The third response message 410 may include an indication of which of the two or more candidate network entities satisfy the one or more second selection criteria. The NWDAF 410 may thus, for example, select the candidate network entities indicated as satisfying the one or more second selection criteria to participate in the collaborative learning process.

[0083] In certain embodiments, the NWDAF 308 receives one or more participation criteria for participating in the collaborative learning process for at least one of the two or more candidate network entities. The one or more participation criteria may be included in the third response message 410 and may have been specified by at least one of the candidate network entities (e.g., in the second response message 408).

[0084] The one or more participation criteria may include any suitable criteria, including, for example, a requirement that a minimum threshold number of network entities participate in the collaborative learning process, a requirement that the NWDAF 308 is in the same network slice as the candidate network entity, and / or a requirement that all network entities participating in the collaborative learning process are in the same network slice.

[0085] The NWDAF 308 may select one or more candidate network entities to participate in the collaborative learning process based on one or more participation criteria. Thus, for example, if the participation criteria specified by a candidate network device are not met, the NWDAF 308 may not select the candidate network device. Conversely, if the one or more participation criteria specified by the candidate network device are met, the NWDAF 308 may select the candidate network device to participate in the collaborative learning process.

[0086] The NWDAF 308 may select a candidate network entity using any of the above methods. The NWDAF 308 may then initiate training of the model using a machine learning algorithm at the selected candidate network entity as part of a collaborative learning process. The NWDAF 308 may send a trigger message to each selected candidate network entity, the trigger message instructing the candidate network entity to train the model using the machine learning algorithm. For example, the trigger message may include information related to the model (e.g., model parameters and / or identification information of the model) and / or an indication of the machine learning algorithm to be used. Alternatively, the candidate network entity may be preconfigured with some or all of this information.

[0087] In About Figure 4 In the described embodiment, the NWDAF queries the candidate network entities by sending a query to an intermediate network entity, in this case, the OAM 312. However, those skilled in the art will appreciate that in some embodiments, the NWDAF 308 may query the candidate network entities directly by sending a second request message 404 to each candidate network entity and receiving one or more responses from one or more candidate network entities. In such an embodiment, the selection of a subset of candidate network entities at the OAM 312 may be omitted.

[0088] Figure 5 is a flow chart of a method performed by a coordination network entity in a communication network according to an embodiment of the present disclosure. For example, the coordination entity may be Figure 3 The NWDAF 302 described. The coordination entity may be in a core network of the communication network. Alternatively, the coordination network may be in a radio access network or a backhaul network of the communication network, for example.

[0089] The method starts at step 502, where the coordination network entity sends a request message for identification information of multiple candidate network entities in the communication network that can perform collaborative learning to a network registration entity in the communication network. The network registration entity may be, for example, in the core network of the communication network. The network registration entity may be, for example, the network registration entity described above. Figure 3 For example, step 502 may substantially correspond to the network function repository function (NRF) described above. Figure 4 The sending of the first request message 400 is described.

[0090] In step 504, the coordination network entity receives identification information of a plurality of candidate network entities from the network registration entity. For example, step 504 may substantially correspond to the above Figure 4 Receipt of a first response message 402 is described.

[0091] In step 506, the coordinating network entity initiates training of the model using a machine learning algorithm at one or more network entities of the plurality of candidate network entities as part of a collaborative learning process.

[0092] Figure 6 is a flow chart of a method performed by a network registration entity in a communication network according to an embodiment of the present disclosure. For example, the network registration may be in a core network of the communication network. The network registration entity may be, for example, Figure 3 Describes the Network Feature Repository feature (NRF).

[0093] The method starts at step 602, where a network registration entity receives a request message from a coordination network entity in the communication network, where the request message requests identification information of multiple candidate network entities in the communication network that can perform collaborative learning. For example, the coordination entity may be Figure 3 The coordination entity may be in the core network of the communication network. For example, step 602 may substantially correspond to the above description of Figure 4 Receipt of a first request message 400 is described.

[0094] In step 604, the network registration entity identifies two or more candidate network entities capable of performing collaborative learning from among a plurality of network entities registered at the network registration entity.

[0095] In step 606, the network registration entity sends an indication of two or more candidate network entities to the coordination network entity. For example, step 606 may substantially correspond to the above description of Figure 4 The sending of the first response message 402 is described.

[0096] Figure 7 is a communication network (for example, Figure 3 Schematic diagram of an apparatus 700 of the system 300 shown in FIG. 700 can be in a network function or entity (e.g., Figure 3 The apparatus 700 is operable to perform the reference Figure 5 The example methods described herein and possibly any other processes or methods disclosed herein. It should also be understood that Figure 5The method may not necessarily be performed only by the apparatus 700. At least some operations of the method may be performed by one or more other entities.

[0097] The apparatus 700 includes processing circuitry 702 (e.g., one or more processors, digital signal processors, general purpose processing units, etc.), machine-readable media 704 (e.g., memory, such as read-only memory (ROM), random access memory, cache memory, flash memory devices, optical storage devices, etc.), and one or more interfaces 706.

[0098] According to an embodiment of the present disclosure, the machine-readable medium 704 stores instructions, which, when executed by the processing circuit 702, cause the apparatus 700 to: send a request message for identification information of a plurality of candidate network entities in the communication network that can perform collaborative learning to a network registration entity in the communication network; receive identification information of the plurality of candidate network entities from the network registration entity; and initiate training of a model using a machine learning algorithm at one or more of the plurality of candidate network entities as part of a collaborative learning process.

[0099] In other embodiments, the processing circuit 702 may be configured to directly perform the method, or configured to cause the apparatus 700 to perform the method without executing instructions stored in the non-transitory machine-readable medium 704, such as by appropriately configured dedicated circuitry.

[0100] One or more interfaces 706 may include hardware and / or software suitable for communicating with other nodes of the communication network using any suitable communication medium. For example, the interface 106 may include one or more wired interfaces, using optical or electrical transmission media. Such an interface may therefore utilize optical or electrical transmitters and receivers, as well as necessary software to encode and decode signals sent via the interface. In another example, the interface 106 may include one or more wireless interfaces. Such an interface may therefore utilize one or more antennas, baseband circuits, etc. The components are shown as coupled together in series; however, those skilled in the art will appreciate that the components may be coupled together in any suitable manner (e.g., via a system bus, etc.).

[0101] In another embodiment of the present disclosure, the device 700 may include a power supply circuit (not shown). The power supply circuit may include a power management circuit or be coupled to a power management circuit and is configured to provide power to the components of the device 700 for performing the functions described herein. The power supply circuit may receive power from a power supply. The power supply and / or the power supply circuit may be configured to provide power to the various components of the device 700 in a form suitable for the various components (e.g., at the voltage and current levels required by each respective component). The power supply may be included in the power supply circuit and / or the device 700 or outside the power supply circuit and / or the device 700. For example, the device 700 may be connected to an external power supply (e.g., an electrical socket) via an input circuit or an interface such as a cable, whereby the external power supply supplies power to the power supply circuit. As another example, the power supply may include a power supply in the form of a battery or a battery pack, which is connected to or integrated in the power supply circuit. If the external power supply fails, the battery may provide backup power. Other types of power supplies, such as photovoltaic devices, may also be used.

[0102] Figure 8 is a communication network (for example, Figure 3 Schematic diagram of an apparatus 800 of the system 300 shown in FIG. 800 can be in a network entity or function (e.g., Figure 3 The apparatus 800 is operable to perform the reference Figure 5 The example methods described herein and possibly any other processes or methods disclosed herein. It should also be understood that Figure 5 The method may not necessarily be performed only by the apparatus 800. At least some operations of the method may be performed by one or more other entities.

[0103] The apparatus 800 includes a sending unit 802 configured to send a request message for identification information of multiple candidate network entities in the communication network that can perform collaborative learning to a network registration entity in the communication network. The apparatus 800 also includes a receiving unit 804 and an initiating unit 806. The receiving unit 804 is configured to receive identification information of multiple candidate network entities from the network registration entity. The initiating unit 806 is configured to initiate training of a model using a machine learning algorithm at one or more network entities among the multiple candidate network entities as part of a collaborative learning process.

[0104] The device 800 may include a processing circuit, which may include one or more microprocessors or microcontrollers and other digital hardware (which may include a digital signal processor (DSP), dedicated digital logic, etc.). The processing circuit may be configured to execute program code stored in a memory, which may include one or more types of memory, such as a read-only memory (ROM), a random access memory, a cache memory, a flash memory device, an optical storage device, etc. In several embodiments, the program code stored in the memory includes program instructions for executing one or more telecommunications and / or data communication protocols and instructions for executing one or more techniques described herein. In some implementations, the processing circuit may be used to cause the sending unit 802, the receiving unit 804, the initiating unit 806, and any other suitable unit of the device 800 to perform corresponding functions according to one or more embodiments of the present disclosure.

[0105] The apparatus 800 may additionally include a power circuit (not shown) configured to provide power to the apparatus 800 .

[0106] Fig. 9 is a communication network (for example, Figure 3 Schematic diagram of an apparatus 900 of the system 300 shown in FIG. 900 can be a network function or entity (e.g., Figure 3 The apparatus 900 is operable to perform a reference Figure 6 The example methods described herein and possibly any other processes or methods disclosed herein. It should also be understood that Figure 6 The method may not necessarily be performed only by the apparatus 900. At least some operations of the method may be performed by one or more other entities.

[0107] The apparatus 900 includes processing circuitry 902 (e.g., one or more processors, digital signal processors, general purpose processing units, etc.), machine-readable media 904 (e.g., memory, such as read-only memory (ROM), random access memory, cache memory, flash memory devices, optical storage devices, etc.), and one or more interfaces 906.

[0108] In one embodiment, the machine-readable medium 904 stores instructions that, when executed by the processing circuit 902, cause the device 900 to: receive a request message from a coordinating network entity in a communication network, the request message requesting identification information of multiple candidate network entities in the communication network that are capable of performing collaborative learning; identify one or more candidate network entities that are capable of performing collaborative learning from multiple network entities registered at the device 900; and send an indication of the one or more candidate network entities to the coordinating network entity.

[0109] In other embodiments, the processing circuit 902 may be configured to directly perform the method, or configured to cause the apparatus 900 to perform the method without executing instructions stored in the non-transitory machine-readable medium 904, such as by appropriately configured dedicated circuitry.

[0110] One or more interfaces 906 may include hardware and / or software suitable for communicating with other nodes of the communication network using any suitable communication medium. For example, the interface 906 may include one or more wired interfaces (using optical or electrical transmission media). Such an interface can therefore utilize optical or electrical transmitters and receivers, as well as necessary software to encode and decode signals sent via the interface. In another example, the interface 906 may include one or more wireless interfaces. Such an interface can therefore utilize one or more antennas, baseband circuits, etc. The components are shown as coupled together in series; however, those skilled in the art will appreciate that the components can be coupled together in any suitable manner (e.g., via a system bus, etc.).

[0111] In another embodiment of the present disclosure, the device 900 may include a power circuit (not shown). The power circuit may include a power management circuit or be coupled to a power management circuit and is configured to provide power to the components of the device 900 for performing the functions described herein. The power circuit may receive power from a power source. The power source and / or the power circuit may be configured to provide power to the various components of the device 900 in a form suitable for the various components (e.g., at the voltage and current levels required by each respective component). The power source may be included in the power circuit and / or the device 900 or outside the power circuit and / or the device 900. For example, the device 900 may be connected to an external power source (e.g., an electrical socket) via an input circuit or an interface such as a cable, whereby the external power source supplies power to the power circuit. As another example, the power source may include a power source in the form of a battery or a battery pack, which is connected to or integrated in the power circuit. If the external power source fails, the battery may provide backup power. Other types of power sources, such as photovoltaic devices, may also be used.

[0112] Fig.10 is a communication network (for example, Figure 3 Schematic diagram of the apparatus 1000 of the system 300 shown in FIG. The apparatus 1000 may be in a network function or entity (eg, Figure 3 The apparatus 1000 is operable to perform a reference Figure 6 The example methods described herein and possibly any other processes or methods disclosed herein. It should also be understood that Figure 6 The method may not necessarily be performed only by the apparatus 1000. At least some operations of the method may be performed by one or more other entities.

[0113] The apparatus 1000 includes a receiving unit 1002 configured to receive a request message from a coordination network entity in a communication network, the request message requesting identification information of multiple candidate network entities in the communication network capable of performing collaborative learning. The apparatus 1000 also includes an identification unit 1004 configured to identify one or more candidate network entities capable of performing collaborative learning from multiple network entities registered at the apparatus. The apparatus 1000 also includes a sending unit 1006 configured to send an indication of one or more candidate network entities to the coordination network entity.

[0114] The device 1000 may include a processing circuit, which may include one or more microprocessors or microcontrollers and other digital hardware (which may include a digital signal processor (DSP), dedicated digital logic, etc.). The processing circuit may be configured to execute program code stored in a memory, which may include one or more types of memory, such as a read-only memory (ROM), a random access memory, a cache memory, a flash memory device, an optical storage device, etc. In several embodiments, the program code stored in the memory includes program instructions for executing one or more telecommunications and / or data communication protocols and instructions for executing one or more techniques described herein. In some implementations, the processing circuit may be used to cause the receiving unit 1002, the identification unit 1004, the sending unit 1006, and any other suitable unit of the device 1000 to perform corresponding functions according to one or more embodiments of the present disclosure.

[0115] The apparatus 1000 may additionally include a power circuit (not shown) configured to provide power to the apparatus 1000 .

[0116] The term "unit" may have a conventional meaning in the field of electronic products, electrical devices and / or electronic devices, and may include, for example, electrical and / or electronic circuits, devices, modules, processors, memories, logical solid-state and / or discrete devices, computer programs or instructions for performing various tasks, processes, calculations, output and / or display functions, etc. (such as those described herein).

[0117] Thus, the embodiments described herein allow for reduced latency, minimized resource overhead, and reduced risk of security issues when implementing machine learning in communication networks. Specifically, the embodiments described herein provide an efficient method for selecting network entities to perform a collaborative learning process, thereby providing a more reliable method for performing collaborative learning in a communication network.

[0118] It should be noted that the above embodiments illustrate rather than limit the concepts disclosed herein, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the attached claims. The word "comprising" does not exclude the presence of elements or steps other than those listed in the claims, "a" or "an" does not exclude a plurality, and a single processor or other unit may perform the functions of several units recorded in the claims. Any reference numerals in the claims should not be construed as limiting their scope.

Claims

1. A method performed by a coordination network entity in a communication network, the method comprising: Sending a first request message from the coordination network entity to a network registration entity in the communication network, for requesting identification information of a plurality of candidate network entities in the communication network that can perform collaborative learning; receiving, at the coordination network entity, identification information of the plurality of candidate network entities from the network registration entity; sending a second request message, the second request message comprising at least one query for additional information of the plurality of candidate network entities; selecting one or more network entities from the plurality of candidate network entities based on the one or more responses to the at least one query; as well as Training a model using a machine learning algorithm is initiated at one or more network entities selected from the plurality of candidate network entities as part of a collaborative learning process.

2. The method according to claim 1, wherein: The coordinating network entity is the Network Data Analysis Function NWDAF.

3. The method according to claim 1, wherein: The network registration entity is the Network Function Repository Function NRF.

4. A coordination network entity for a communication network, the coordination network entity comprising a processing circuit and a non-transitory machine-readable medium storing instructions, the instructions, when executed by the processing circuit, causing the coordination network entity to: Sending a first request message from the coordination network entity to a network registration entity in the communication network, for requesting identification information of a plurality of candidate network entities in the communication network that can perform collaborative learning; receiving, at the coordination network entity, identification information of the plurality of candidate network entities from the network registration entity; sending a second request message, the second request message comprising at least one query for additional information of the plurality of candidate network entities; selecting one or more network entities from the plurality of candidate network entities based on the one or more responses to the at least one query; as well as Training a model using a machine learning algorithm is initiated at one or more network entities selected from the plurality of candidate network entities as part of a collaborative learning process.

5. The coordination network entity according to claim 4, wherein: The request message includes one or more criteria for selecting a candidate network entity to perform the collaborative learning process.

6. The coordination network entity according to claim 5, wherein: The one or more criteria include one or more of the following: at least one primary criterion related to the ability of the candidate network entity to perform the collaborative learning process; and At least one secondary criterion related to an ability of the candidate network entity to respond to the query type.

7. The coordination network entity according to claim 4, wherein: The at least one query for additional information relates to one or more of the following: configuration of the candidate network entity; performance requirements for the candidate network entity; the availability of training data at the candidate network entity for training the model; and Attributes of the training data available at the candidate network entity.

8. The coordination network entity according to claim 4, wherein: One or more of the following applies: The coordinating network entity is a network data analysis function NWDAF; and The network registration entity is the Network Function Repository Function NRF.

9. The coordination network entity according to claim 4, wherein: One or more of the coordination network entity and the network registration entity are in a core network of the communication network.

10. A network registration entity for a communication network, the network registration entity comprising a processing circuit and a non-transitory machine-readable medium storing instructions which, when executed by the processing circuit, cause the network registration entity to: receiving, at the network registration entity, a request message from a coordinating network entity in the communication network, the request message comprising one or more criteria for selecting a candidate network entity from a plurality of network entities to train a model using a machine learning algorithm as part of a collaborative learning process; identifying two or more candidate network entities satisfying the one or more criteria from among a plurality of network entities registered with the network registration entity; storing capability information for each of the plurality of candidate network entities registered at the network registration entity, the capability information comprising an indication of whether the network entity is configured to respond to a query type; and An indication regarding the two or more candidate network entities is sent from the network registration entity to the coordinating network entity.

11. The network registration entity according to claim 10, wherein: The one or more criteria include one or more of the following: at least one primary criterion related to the candidate network entity's ability to perform a collaborative learning process; and At least one secondary criterion related to an ability of the candidate network entity to respond to the query type.

12. The network registration entity according to claim 11, wherein: The query types include one or more of the following: queries related to the configuration of the candidate network entity; a query related to performance requirements for the candidate network entity; a query related to availability of training data at the candidate network entity for training a model; as well as A query related to attributes of the training data available at the candidate network entity.

13. The network registration entity according to claim 10, wherein: The stored capability information is stored in a profile of each network entity of the plurality of network entities registered at the network registration entity.

14. The network registration entity according to claim 13, wherein: The stored capability information also includes, for each network entity, an indication of whether the network entity is capable of performing collaborative learning.

15. The network registration entity according to claim 10, wherein: One or more of the following applies: The network registration entity is a Network Function Repository Function NRF; and The coordinating network entity is the Network Data Analysis Function NWDAF.

16. The network registration entity according to claim 10, wherein: One or more of the network registration entity and the coordination network entity are in a core network of the communication network.

17. A system in a communication network, the system comprising a coordination network entity and a network registration entity, wherein: The network registration entity includes a first processing circuit and a first non-transitory machine-readable medium storing instructions that, when executed by the first processing circuit, cause the network registration entity to: receiving, at the network registration entity, a request message from the coordination network entity, the request message requesting identification information of a plurality of candidate network entities in the communication network capable of performing collaborative learning; identifying two or more candidate network entities capable of performing collaborative learning from a plurality of network entities registered at the network registration entity; storing capability information for each of the plurality of candidate network entities registered at the network registration entity, the capability information comprising an indication of whether the network entity is configured to respond to a query type; and The coordination network entity includes a second processing circuit and a second non-transitory machine-readable medium storing instructions, wherein the instructions, when executed by the second processing circuit, cause the coordination network entity to: receiving, at the coordinating network entity, identification information of the two or more candidate network entities from the network registration entity, and At one or more of the two or more candidate network entities, training of a model using a machine learning algorithm is initiated as part of a collaborative learning process.

18. The system of claim 17, wherein: The one or more criteria include at least one primary criterion related to the ability of the candidate network entity to perform a collaborative learning process.

19. The system of claim 17, wherein the query type comprises one or more of the following: a query relating to the availability of training data at the candidate network entity for training a model; and A query related to attributes of the training data available at the candidate network entity.

Images