Description

Arrangement of Communication Network Components and Method for Providing a Machine Learning Model for Performing Communication Network Analytics

The present disclosure relates to arrangements of communication network components and methods for providing a machine learning model for performing communication network analytics.

In a communication network such as a 5G mobile communication network it is important to ensure that a certain service quality can be maintained. For this, network statistical analysis and prediction information which may be generated from information like load, resource usage, available components, state of components etc. may be monitored to be able to take measures, e.g. when overload is imminent, to avoid a drop of service quality, etc. In a communication network having a network slice architecture, like a 5G mobile communication system, this relates in particular to monitoring of network analysis and prediction information (also referred to as network analytics information) for individual network slice instances since, for example, individual network slice instances may become overloaded. Accordingly, efficient and reliable approaches to provide network analysis and/or prediction information are desirable.

According to one aspect of the present invention an arrangement of communication network components of a mobile communication network is provided comprising a network analytics component configured to send a request for information about a machine learning model trained to derive network analytics information, wherein the request includes a specification of a quality requirement that should be fulfilled by the machine learning model; and an information providing component configured to determine and provide the requested information to the network analytics component in response to the request; wherein the information providing component is a Network Repository Function or an Operation, Administration and Maintenance system, wherein the information providing component is a machine learning model training component and the requested information is a specification of the machine learning model allowing implementation of the machine learning model by the network analytics component; wherein the machine learning model training component is configured to train a plurality of machine learning models for deriving network analytics information and is configured to select a machine learning model from the plurality of machine learning models which fulfils the quality requirement and provide a specification of the selected machine learning model to the network analytics component allowing implementation of the machine learning model by the network analytics component.

According to another aspect of the present invention, a method for providing a machine learning model for performing communication network analytics according to the above communication network arrangement is provided.

In the drawings, like reference characters generally refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead generally being placed upon illustrating the principles of the invention. In the following description, various aspects are described with reference to the following drawings, in which:

Figure 1 shows a radio communication system, for example configured according to 5G (Fifth Generation) as specified by 3GPP (Third Generation Partnership Project).

Figure 2 shows a network slice.

Figure 3 shows an NWDAF AnLF (Network Data Analytics Function Analytics Logical Function) and an NWDAF MTLF (Network Data Analytics Function Model Training Logical Function).

Figure 4 shows a flow diagram illustrating an ML (machine learning) model provisioning service operation. Figure 5 shows a flow diagram illustrating an ML model request service operation.

Figure 6 shows an exemplary scenario in which an MTLF provides machine learning (ML) models to a first AnLF and a second AnLF.

Figure 7 shows a flow diagram for a registration of an MTLF at an NRF (Network Repository Function).

Figure 8 shows a flow diagram for discovery of an MTLF by an NWDAF service consumer by consulting an NRF.

Figure 9 shows a communication network arrangement of a mobile communication network according to an embodiment.

Figure 10 shows a flow diagram illustrating a method for performing communication network analytics according to an embodiment.

The following detailed description refers to the accompanying drawings that show, by way of illustration, specific details and aspects of this disclosure in which the invention may be practiced. Other aspects may be utilized and structural, logical, and electrical changes may be made without departing from the scope of the invention. The various aspects of this disclosure are not necessarily mutually exclusive, as some aspects of this disclosure can be combined with one or more other aspects of this disclosure to form new aspects.

Various examples corresponding to aspects of this disclosure are described below:

Example 1 is an arrangement of communication network components of a mobile communication network including a network analytics component configured to send a request for information about a machine learning model trained to derive network analytics information, wherein the request includes a specification of a quality requirement that should be fulfilled by the machine learning model and an information providing component configured to determine and provide the requested information to the network analytics component in response to the request.

Example 2 is the arrangement of communication network components of Example 1, wherein the request is a discovery request and the requested information is an identification of a machine learning model training component providing a machine learning model fulfilling the quality requirement and wherein the network analytics component is configured to request the machine learning model from the machine learning model training component using the identification of the machine learning model training component.

Example 3 is the arrangement of communication network components of Example 2, wherein the information further includes an identification of the machine learning model fulfilling the quality requirement and the network analytics component is configured to request the machine learning model from the machine learning model training component using the identification of the machine learning model.

Example 4 is the arrangement of communication network components of Example 2 or 3, wherein the information providing component is a network component storing profiles of network function components of the mobile communication network.

Example 5 is arrangement of communication network components of any one of Examples 1 to 4, wherein the information providing component is a Network Repository Function or an Operation, Administration and Maintenance system.

Example 6 is the arrangement of communication network components of Example 1, wherein the information providing component is a machine learning model training component and the requested information is a specification of the machine learning model allowing implementation of the machine learning model by the network analytics component.

Example 7 is the arrangement of communication network components of Example 6, wherein the machine learning model training component is configured to train a plurality of machine learning models for deriving network analytics information and is configured to select a machine learning model from the plurality of machine learning models which fulfils the quality requirement and provide a specification of the selected machine learning model to the network analytics component allowing implementation of the machine learning model by the network analytics component.

Example 8 is the arrangement of communication network components of any one of Examples 1 to 7, wherein the quality requirement is a required accuracy and/or a required efficiency and/or a required interoperability between a plurality of network analytics components.

Example 9 is the arrangement of communication network components of Example 8, wherein the required interoperability between a plurality of network analytics components is the requirement that the machine learning model is operable with the network analytics component, wherein the network analytics component is provided by a different service provider than a machine learning model training component service provider of a machine learning model training component generating the machine learning model.

Example 10 is the arrangement of communication network components of any one of Examples 1 to 9, wherein the quality requirement is a support of a type of the network analytics component of multiple network analytics component types.

Example 11 is a method for providing a machine learning model for performing communication network analytics including sending a request for information about a machine learning model trained to derive network analytics information, wherein the request includes a specification of a quality requirement that should be fulfilled by the machine learning model and determining and providing the requested information in response to the request.

It should be noted that one or more of the features of any of the examples above may be combined with any one of the other examples. In particular, the Examples described in context of the communication network arrangement are analogously valid for the method.

According to a further aspect of the present invnetion, a computer program and a computer readable medium including instructions, which, when executed by a computer, make the computer perform the method of any one of the above Examples are provided.

In the following, various examples will be described in more detail.

Figure 1 shows a mobile radio communication system 100, for example configured according to 5G (Fifth Generation) as specified by 3GPP (Third Generation Partnership Project).

The mobile radio communication system 100 includes a mobile radio terminal device 102 such as a UE (user equipment), a nano equipment (NE), and the like. The mobile radio terminal device 102, also referred to as subscriber terminal, forms the terminal side while the other components of the mobile radio communication system 100 described in the following are part of the mobile communication network side, i.e. part of a mobile communication network (e.g. a Public Land Mobile Network PLMN).

Furthermore, the mobile radio communication system 100 includes a Radio Access Network (RAN) 103, which may include a plurality of radio access network nodes, i.e. base stations configured to provide radio access in accordance with a 5G (Fifth Generation) radio access technology (5G New Radio). It should be noted that the mobile radio communication system 100 may also be configured in accordance with LTE (Long Term Evolution) or another mobile radio communication standard (e.g. non-3GPP accesses like WiFi) but 5G is herein used as an example. Each radio access network node may provide a radio communication with the mobile radio terminal device 102 over an air interface. It should be noted that the radio access network 103 may include any number of radio access network nodes.

The mobile radio communication system 100 further includes a core network (5GC) 119 including an Access and Mobility Management Function (AMF) 101 connected to the RAN 103, a Unified Data Management (UDM) 104 and a Network Slice Selection Function (NSSF) 105. Here and in the following examples, the UDM may further consist of the actual UE’s subscription database, which is known as, for example, the UDR (Unified Data Repository). The core network 119 further includes an AUSF (Authentication Server Function) 114 and a PCF (Policy Control Function) 115.

The core network 119 may have multiple core network slices 106, 107 and for each core network slice 106, 107, the operator (also referred to MNO for Mobile Network Operator) may create multiple core network slice instances 108, 109. For example, the core network 119 includes a first core network slice 106 with three core network slice instances (CNIs) 108 for providing Enhanced Mobile Broadband (eMBB) and a second core network slice 107 with three core network slice instances (CNIs) 109 for providing Vehicle-to-Everything (V2X).

Typically, when a core network slice is deployed (i.e. created), network functions (NFs) are instantiated, or (if already instantiated) referenced to form a core network slice instance and network functions that belong to a core network slice instance are configured with a core network slice instance identification.

Specifically, in the shown example, each instance 108 of the first core network slice 106 includes a first Session Management Function (SMF) 110 and a first User Plane Function (UPF) 111 and each instance 109 of the second core network slice 107 includes a second Session Management Function (SMF) 112 and a second User Plane Function (UPF) 113. The SMFs 110, 112 are for handling PDU (Protocol Data Unit) sessions, i.e. for creating, updating and removing PDU sessions and managing session context with the User Plane

Function (UPF).

The RAN 103 and the core network 119 form the network side of the mobile radio communication system, or, in other words, form the mobile radio communication network. The mobile radio communication network and the mobile terminals accessing the mobile radio communication network form, together, the mobile radio communication system.

Like the core network 119, the RAN 103 may also be sliced, i.e. include multiple RAN slices. A RAN slice and a core network slice 106, 107 may be grouped to form a network slice as illustrated in figure 2.

Figure 2 shows a network slice 200.

A network slice 200 can be seen as a logical network that provides specific network capabilities and network characteristics.

The network slice 200 includes a RAN slice 201 and a core network (e.g. 5GC) slice 202.

An instance of a network slice 200, i.e. a network slice instance which includes a RAN slice instance and a core network slice instance, is defined within a PLMN and includes Core Network Control Plane and User Plane Network Functions, RAN components (e.g. NG-RAN components), in particular base stations. In case of roaming these components may at least partially be of a visited PLMN (VPLMN).

In the following, “network slice” (or just “slice”) generally refers to a core network slice but may also include a RAN slice.

An S-NSSAI (Single Network Slice Selection Assistance information) identifies a network slice and consists of:

- A Slice/Service type (SST), which refers to the expected Network Slice behaviour in terms of features and services;

- A Slice Differentiator (SD) which is optional information that complements the slice/service type(s) to differentiate amongst multiple network slices of the same slice/service type.

A network slice instance is identified by an NSI (also referred to as NSI ID).

NSSAI may include one or more S-NSSAIs.

Allowed NSSAI is NSSAI provided by the serving PLMN (Public Land Mobile Network) during e.g. a registration procedure, indicating the S-NSSAI values allowed by the network for a UE in the serving PLMN for the current registration area.

Configured NSSAI is NSSAI that has been provisioned in the UE. It may be applicable to one or more PLMNs.

Requested NSSAI is NSSAI that the UE provides to the network during registration in order to establish a PDU session to the requested network slice.

The mobile radio communication system 100 may further include an 0AM (Operation, Administration and Maintenance) function (or entity) 116, e.g. implemented by one or more 0AM servers which is connected to the RAN 103 and the core network 119 (connections are not shown for simplicity). The 0AM 116 may include an MDAS (Management Data Analytics Service). The MDAS may for example provide an analytics report regarding network slice instance load. Various factors may impact the network slice instance load, e.g. number of UEs accessing the network, number of QoS flows, the resource utilizations of different NFs which are related with the network slice instance.

Further, the core network 118 includes an NRF (Network Repository Function).

The core network 119 may further include a Network Data Analytics Function (NWDAF) 117. The NWDAF is responsible for providing network analysis and/or prediction information upon request from network functions. For example, a network function may request specific analysis information on the load level of a particular network slice instance. Alternatively, the network function can use the subscribe service to ensure that it is notified by the NWDAF if the load level of a network slice instance changes or reaches a specific threshold. The NWDAF 117 may have an interface to various network functions on the mobile communication network side, e.g. to the AMF 101, the SMFs 110, 112 and the PCF 115. For simplicity, only the interface between the NWDAF 117 and the AMF 101 is depicted.

For example, NWDAF analytics should allow monitoring the number of UEs registered in a network slice instance and their Observed Service Experience. In addition to 0AM performing SLA (service level agreement) assurance, 5GC NFs may take actions based on NWDAF slice QoE analytics to prevent further service experience degradation in the network slice instance.

The NSSF 105 or AMF 101 may for example determine when a load distribution decision is required to address an issue identified by processing the analytics result (i.e. network analysis and/or prediction information) provided by the NWDAF 117. For example, when a network slice instance is detected or forecast to experience service experience degradation, new UE registrations or PDU sessions may not be assigned to that network slice instance anymore by triggering a network slice load distribution mechanism. The NSSF 105, AMF 101 and/or 0AM 116, for example, may also simultaneously subscribe to both slice service experience and slice load analytics from the NWDAF 117. One or multiple subscriptions to one or multiple S- NSSAI(s) and NSI(s) are possible.

The NWDAF 117 provides slice load level information to an NF on a Network slice instance level. The NWDAF notifies slice specific network status analytics information (also referred to as network analysis and/or prediction information) to the NFs that are subscribed to it. For example, as network analysis and/or prediction information, the NWDAF 117 reports when the load level of the Network slice instance, indicated by the S-NSSAI and the associated NSI (if applicable) in the Analytics Filter, crosses the threshold provided in the analytics subscription; if no threshold is provided in the subscription, the reporting (Notify operation) is assumed to be periodic.

To generate the network analysis and/or prediction information, NWDAF 117 collects input data required (e.g. to derive slice service experience analytics), i.e. information for analysis of the state of the network slice instance. The NWDAF 117 may obtain such kind of information by subscribing to network functions to be informed accordingly.

According to 3GPP Release 17 (Rel-17) the NWDAF 117 is decomposed into two functions.

Figure 3 shows an NWDAF AnLF (Analytics Logical Function) 301 and an NWDAF MTLF (Model Training Logical Function) 302.

An NWDAF 301 containing the Analytics logical function is denoted as NWDAF(AnLF) or NWDAF-AnLF or simply AnLF and can perform inference, derive analytics information and expose analytics service i.e. Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo. An NWDAF 302 containing the Model Training logical function is denoted as NWDAF(MTLF) or NWDAF-MTLF or simply MTLF, trains machine learning (ML) models and exposes new training services (e.g. providing trained model).

The NWDAF(AnLF) 301 provides an analytics service. It may be contacted by an NF (acting as service consumer) 303 (e.g. an AMF or SMF) to be provided with analytics information.

Input request parameters to the NWDAF(AnLF) 301 by the NF 303 are, e.g. an Analytic ID and an S-NSSAI. The output from the NWDAF (AnLF) 301 are for example statistics and/or prediction (referred to as network analytics information herein) for an Analytic ID.

Examples of Analytics IDs are UE communication, UE mobility, UE behaviour, User data congestion and Network performance, etc.

The NWDAF(MTLF) 302 provides an ML model service (Model Training Exposure). It may be contacted by another NWDAF (acting as service consumer such as an AnLF) 304 to be provided with a trained machine learning model to perform analytics.

Table 1 lists ML model related service operations.

The Nnwdaf_MLModelProvision service and the Nnwdaf_MLModelInfo service are provided by an NWDAF(MTLF) 301 and consumed by an NWDAF 304 (which is usually an NWDAF(AnLF)).

Figure 4 shows a flow diagram 400 illustrating an ML model provisioning service operation.

An NWDAF service consumer 401 (e.g. corresponding to the NWDAF 304 or AnLF 301) and an NWDAF (MTLF) 402 (e.g. corresponding to NWDAF (MTLF) 302) are involved in the flow.

The purpose of the flow of figure 4 is for the NWDAF service consumer 401 to subscribe and/or to be notified when ML model information on the related analytics becomes available.

In 403, the NWDAF service consumer 401 subscribes to the model provisioning service (it may later similarly unsubscribe). The NWDAF(MTL) 402 acknowledges in 404.

When subscribing, the NWDAF service consumer 401 may provide the following input for the NWDAF(MTLF) 402:

• A list of Analytics ID(s): identifies the analytics for which the ML model is to be used.

• Filter Information: indicates the conditions to be fulfilled for reporting Analytics Information. This enables to select which ML model for the analytics is requested.

• Target of Analytics Reporting: indicates the object(s) for which ML model for the analytics is requested, entities such as specific UEs, a group of UE(s) or any UE (i.e. all UEs).

When the NWDAF service consumer 401 has subscribed to the model provisioning service, the NWDAF(MTLF) provides as output notification correlation information and ML model information (e.g. ML model file address(es) for the Analytics ID(s)) to the NWDAF service consumer 401.

It should be noted that an NWDAF 117 can provide the analytics service or the ML model service or both services (i.e. an NWDAF 117 may be both an NWDAF(AnLF) 301 and a NWDAF(MTLF) 302).

Figure 5 shows a flow diagram 500 illustrating an ML model request service operation.

An NWDAF service consumer 501 (e.g. corresponding to the NWDAF 304, e.g. an AnLF) and an NWDAF(MTLF) 502 (e.g. corresponding to NWDAF(MTLF) 302) are involved in the flow.

The purpose of the flow of figure 5 is that the NWDAF service consumer 501 is provided with (a set of) ML model(s) information associated with Analytics ID, so that the NWDAF service consumer 501 can derive analytics.

In 503, the NWDAF service consumer 501 sends a request message to the NWDAF(MTLF) 502 having similar input as in the flow of figure 4 (in particular a one or more Analytic IDs) to request NWDAF ML model information. This input optionally includes Filter information to indicate the conditions to be fulfilled for reporting Analytics Information, and Target of Analytics Reporting to indicate the object(s) for which ML model for the analytics is requested, entities such as specific UEs, a group of UE(s) or any UE (i.e. all UEs).

In 504, the NWDAF(MTLF) 502 responds with a similar output as in the flow of figure

4, i.e., one or more tuples (Analytics ID, address (e.g. URL (Uniform Resource Locator) or FQDN (Fully-Qualified Domain Name)) of Model file or Analytical Data Repository Function address/ID).

The NWDAF(MTLF) 302 may determine the machine learning model to be provided for the analytics depending on whether the ML model should be provided for analysis of specific UEs or a group of UE(s) or any UE (i.e. all UEs).

So, using the flows of FIG. 4 and FIG. 5 a service consumer (of an MTLF provision service), e.g. an AnLF, i.e. generally a network analytics component, can acquire a machine learning model (for deriving network analytics information) from a provider of the service (i.e. an MTLF).

There may be multiple service consumers who request machine learning models from an MTLF, as it is illustrated in figure 6.

Figure 6 shows an exemplary scenario in which an MTLF 601 provides machine learning (ML) models (i.e. specifications of machine learning models such that they may be implemented, e.g. a specification of weights of a neural network) to a first AnLF 602 and a second AnLF 603.

The AnLFs 602 and 603 may have different requirements with respect to the ML models they require. For example, they may be from different providers (e.g. vendors or network function service provider) and thus have different functionalities. Further, the MTLF 601 and each of the AnLFs 602, 603 may be from different providers so the MTLF 601 may not have knowledge about the requirements of the AnLFs 602, 603.

According to various embodiments, approaches are described which enable sharing ML models in such a scenario, for example in a case the MTLF 601 is provided by a first provider (e.g. vendor), and the AnLFs 602, 603 are provided by a second and a third provider, respectively.

It is also possible that the MTLF 601 specifically designs an ML model for the first AnLF 602 (e.g. because the MTLF 601 and the first AnLF 602 are from the same provider). In particular in such a case, the second AnLF 603 may require an ML model which is interoperable, i.e. which can also be used by the AnLF 603 (which may be of another type than the first AnLF 602 and may have different requirements). The interoperability of the ML model, i.e. the capability to be used by different types of AnLFs, may be seen as a quality of the ML model. Accordingly, according to various embodiments, a mechanism (i.e. a functionality, e.g. provided by a certain (signalling) protocol between the respective AnLF 603 and the MTLF 601) is provided that allows the AnLF 603 to acquire an ML model from the MTLF 601 having a certain quality.

According to various embodiments, this is achieved by that the AnLF 603 sends a request message to the MTLF 603 which includes an indication of a quality requirement regarding a machine learning model and/or retrieves the ID of a ML model, e.g. from an NRF, fulfilling the quality requirement.

For example, input parameters included in the request message sent in 403 for subscription to the Nnwdaf MLModelProvision service and input parameters included in the request message sent in 503 for requesting the NnwdafJMLModellnfo service (i.e. to request NWDAF ML model information) may include one or more of the following:

• an indication of a AnLF type that should be supported by the ML model (e.g. an identification of a vendor (e.g., Vendor-ID or Supported Vendor-ID AVP (Attribute value pair))

• an indication that the ML model should be interoperable

• an indication of an accuracy (i.e. accuracy of inference) of the ML model

• an indication of an efficiency of the ML model (i.e., resource cost for using the ML model). The MTLF 603 (e.g. MTLF 402 or MTLF 502) may, in response to the request, provide the ML model and information about inference accuracy (i.e., accuracy of inference) and efficiency (i.e., resource cost for using the ML model) and/or other special treatment required for the ML model. In case the request does not directly request the ML model, the MTLF 603 (i.e. generally the machine learning model training component) may provide an identification of a suitable model (having the required quality) so that the AnLF 603 can directly request it by providing the identification of the ML model, e.g. by including the identification in the request of 403 or the request of 503. In particular, the MTLF 603 may provide multiple identification of ML models (ML IDs) and the AnLF 603 can select and request one of them by sending a corresponding requested including the identification of the selected ML model. As described further below, the AnLF 603 may also acquire an ML ID from an NRF to retrieve the corresponding ML model from an MTLF 601 by specifying the ML ID (whose MTLF ID may also be provided by the NRF).

According to various embodiments, for discovery of an MTLF by a service consumer, e.g. MTLF 601 by the second AnLF 603, the MTLF registers at an NRF as illustrated in figure 7 which may then be consulted by a service consumer as illustrated in figure 8.

Figure 7 shows a flow diagram 700 for a registration of an MTLF 701 at an NRF 702.

In 703, the MTLF sends an registration request to the NRF 702 for requesting registration of the NWDAF(MTLF)’s NF profile.

The NRF 702 registers the MTLF’s NF profile and acknowledges in 704.

Multiple MTLFs can register with the NRF 702 in this manner. For each MTLF, the NF profile includes one or more of the following:

• Analytic ID(s) of analytics supported

• S-NSSAI(s) of supported network slice(s)

Indications of supported areas of interest Identifications of ML models it provides

• Indication of a quality of ML models provided (e.g. identifications of supported ML model types, e.g. supported vendors).

• Identification of supported interoperable ML models

It should be noted that an OAM 116 can also configure the above information in the NRF 702 and can provide the above information to the service consumer (instead of the NRF 702). For example, the OAM 116 can configure the above information in each AnLF along with one or more MTLF addresses. So in this case, the AnLF uses the local configuration, instead of checking with the NRF 702.

Figure 8 shows a flow diagram 800 for discovery of an MTLF by an NWDAF service consumer 801 by consulting an NRF 802.

In 803, the NWDAF service consumer 801 (e.g. an AnLF) sends a discovery request to the NRF 802.

In 804, the NRF 802 responds with a discovery response.

For discovering an MLTF which provides a certain machine learning model or a certain machine learning model of a certain quality, the NWDAF service consumer 801 may include, corresponding to the information included in the NF profile of the MLTF as given above, one or more of

• An Analytic ID(s) of analytics that should be supported by an ML model provided by the MLTF

• S-NSSAI(s) of network slice(s) that should be supported by an ML model provided by the MLTF

Indications of one or more areas of interest that should be supported by an ML model provided by the MLTF

• Identification(s) of one or more ML models that should be provided by the MLTF

• Indication of a quality (or indications of multiple qualities) that should be supported by an ML model provided by the MLTF provided (e.g. identifications of supported ML model types, e.g. supported vendors)

• Identification of interoperability supporting ML models provided by the MTLF

According to the parameters included in the request of 803, the NRF 802 generates a list of (at least one) MLTFs which fulfil the request (e.g. provides the one or more identified ML models or provide one or more ML models having the indicated quality) and sends the list to the NRF 802 in the response of 804. The NRF 802 may also provide supported ML model IDs of ML models provided by the MLTFs. The NRF 802 may also include accuracy and efficiency levels of the ML models provided by the MLTFs of the list.

Based on the NRF response the service consumer 801 selects an MTLF (e.g. for requesting an ML model according to figures 4 and 5).

In summary, according to various embodiments, a communication network is provided as illustrated in figure 9.

Figure 9 shows a communication network arrangement 900 of a mobile communication network according to an embodiment.

The communication network arrangement 900 includes a network analytics component 901 configured to send a request 903 for information 904 about a machine learning model trained to derive network analytics information, wherein the request includes a specification of a quality requirement that should be fulfilled by the machine learning model.

Further, the communication network arrangement 900 includes an information providing component 902 configured to determine and provide the requested information 904 to the network analytics component in response to the request.

The information providing component 902 may provide an ID of a ML model for the network analytics component 901 to retrieve the ML model from a machine learning model training component (which may also be part of the communication network arrangement) or the information providing component 902 may directly provide a complete specification of ML model (e.g. may itself be a machine learning model training component), i.e. a specification allowing implementation of the machine learning model by the network analytics component (e.g. values of the parameters (typically the weights) of the ML model required to implement it).

According to various embodiments, in other words, a network analytics component may request and be provided with a ML model that is suitable for it.

Embodiments as described above of the approach of figure 10 enable an MTLF (in general a machine learning model training component) to register itself with the capability indicating whether the ML model provided by MTLF supports a certain quality (e.g. one or more specific AnLF types.). They allow the NF consumer (in general a network analytics component) to discover the ML model that fulfils the required quality, e.g. fits the NF consumer, i.e., ML model that fits to a specific AnLF type. Based on the required quality, the MTLF can filter the model(s) that fulfil the quality (e.g. are interoperable with certain AnLF types or fits a required accuracy level) and provide it to the NF consumer.

Embodiments thus for example lock-in of deployments in which both AnLF and MTLF are from the same provider (e.g. vendor) and allow AnLF and MTLF to work in a multiprovider environment. Various embodiments introduce input and output parameters to discover and select an MTLF in a multi-provider environment, and, by introducing corresponding logic into the MTLF, to select the a ML model suitable for a consumer (e.g. an interoperable ML model) and provide it to the consumer (e.g. AnLF).

The communication network arrangement 900 for example carries out a method as illustrated in figure 10.

Figure 10 shows a flow diagram 1000 illustrating a method for performing communication network analytics according to an embodiment.

In 1001, a request for information about a machine learning model trained to derive network analytics information is sent, wherein the request includes a specification of a quality requirement that should be fulfilled by the machine learning model.

In 1002, requested information is determined and provided (to the sender of the request) in response to the request

According to various embodiments, a communication network arrangement of a mobile communication network is provided including a machine learning model training component configured to train a plurality of machine learning models for deriving network analytics information, a network analytics component configured to discover the machine learning model training component from repository wherein the machine learning model training component can provide machine learning models to different set of network analytics components, wherein the network analytics component is configured to request a machine learning model from the machine learning model training component, wherein the network analytics component specifies a model parameter (which may for example be a quality) for which it requests the machine learning model. The machine learning model training component is configured to determine the network analytics component set and select a machine learning model from the plurality of machine learning models which has been trained for a model parameters according to the parameter specified in the request and to provide the determined machine learning model to the network analytics component.

The model parameter includes a model ID(s) and/or an interoperability indication (e.g., includes vendor-ID or supported vendor-ID AVP ) and/or indicates accuracy (e.g., accuracy of inference) and/or efficiency (e.g., resource cost for using the ML model).

The method may be performed and the components of the communication network arrangement may for example be implemented by one or more circuits. A "circuit" may be understood as any kind of a logic implementing entity, which may be special purpose circuitry or a processor executing software stored in a memory, firmware, or any combination thereof. Thus a "circuit" may be a hard-wired logic circuit or a programmable logic circuit such as a programmable processor, e.g. a microprocessor. A "circuit" may also be a processor executing software, e.g. any kind of computer program Any other kind of implementation of the respective functions described above may also be understood as a "circuit".

While specific aspects have been described, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the aspects of this disclosure as defined by the appended claims. The scope is thus indicated by the appended claims and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced.