Switching machine learning functionality based on resource availability

Abstract

Systems, methods, apparatuses, and computer program products for switching machine learning functions based on resource availability. A method may include executing a machine learning function, a feature, or a model at a first network element. The method may also include monitoring, at the first network element, machine learning resources that affect performance of the machine learning function, feature, or model. The method may further include indicating to the second network element, based on the monitoring, machine learning related resource availability that affects performance of the machine learning function, feature, or model. Further, the method may include receiving, from a third network element, a configuration of an action to be performed by the first network element based on machine learning related resource availability that affects performance of the machine learning function, feature, or model. Further, the method may include performing an action.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Application No. 63 / 531,685, filed on August 9, 2023, which is incorporated by reference in its entirety into this application. Technical Field

[0003] Some exemplary embodiments may generally relate to mobile or wireless telecommunication systems, such as Long Term Evolution (LTE) or fifth generation (5G) New Radio (NR) access technology, or above 5G, or other communication systems. For example, certain exemplary embodiments may relate to apparatus, systems, and / or methods for switching machine learning functions based on resource availability. Background Art

[0004] Examples of mobile or wireless telecommunication systems may include Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access Network (UTRAN), LTE Evolved UTRAN (E-UTRAN), Advanced LTE (LTE-A), MulteFire, LTE-A Pro, fifth generation (5G) radio access technology or NR access technology, and / or Advanced 5G. 5G wireless systems refer to next generation (NG) radio systems and network architectures. 5G network technology is primarily based on NR technology, but 5G (or NG) networks can also be built on E-UTRAN radios. It is estimated that NR can provide bit rates of about 10-20Gbit / s or higher, and can support at least enhanced mobile broadband (eMBB) and ultra-reliable low latency communications (URLLC) and massive machine type communications (mMTC). NR is expected to deliver extreme broadband and ultra-robust, low latency connectivity and massive networking to support IoT. Summary of the invention

[0005] Some exemplary embodiments may be directed to a method. The method may include executing a machine learning function, feature, or model at a first network element. The method may also include monitoring a machine learning resource at the first network element that affects the performance of the machine learning function, feature, or model. The method may further include indicating to a second network element, based on the monitoring, the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. In addition, the method may include: receiving from a third network element a configuration of an action to be performed by the first network element based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the method may include performing an action.

[0006] Other exemplary embodiments may be directed to a device. The device may include at least one processor and at least one memory including computer program code. The at least one memory and the computer program code may be configured to, together with the at least one processor, cause the device to perform a machine learning function, feature, or model at least at the device. The device may also be enabled to monitor machine learning resources at the device that affect the performance of the machine learning function, feature, or model. The device may further be enabled to indicate to a first network element the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model based on the monitoring. In addition, the device may be enabled to receive from a second network element a configuration of an action to be performed by the device based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the device may be enabled to perform an action.

[0007] Other exemplary embodiments may be directed to a device. The device may include a component for performing a machine learning function, feature, or model at the device. The device may also include a component for monitoring machine learning resources at the device that affect the performance of the machine learning function, feature, or model. The device may further include a component for indicating to a first network element, based on the monitoring, the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. In addition, the device may include a component for receiving from a second network element a configuration of an action to be performed by the device based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the device may include a component for performing an action.

[0008] According to other exemplary embodiments, a non-transitory computer-readable medium may be encoded with instructions that, when executed in hardware, perform a method. The method may include executing a machine learning function, feature, or model at a first network element. The method may also include monitoring machine learning resources that affect the performance of the machine learning function, feature, or model at the first network element. The method may further include indicating, based on the monitoring, to a second network element the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. In addition, the method may include: receiving from a third network element a configuration of an action to be performed by the first network element based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the method may include performing an action.

[0009] Other exemplary embodiments may be directed to a computer program product for performing a method. The method may include executing a machine learning function, feature, or model at a first network element. The method may also include monitoring a machine learning resource at the first network element that affects the performance of the machine learning function, feature, or model. The method may further include indicating to a second network element the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model based on the monitoring. In addition, the method may include: receiving from a third network element a configuration of an action to be performed by the first network element based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the method may include performing an action.

[0010] Other exemplary embodiments may be directed to an apparatus that may include circuitry configured to perform a machine learning function, feature, or model at the apparatus. The apparatus may also include circuitry configured to monitor machine learning resources at the apparatus that affect the performance of the machine learning function, feature, or model. The apparatus may further include circuitry configured to indicate, based on the monitoring, to a first network element the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. In addition, the apparatus may include circuitry configured to receive from a second network element a configuration of an action to be performed by the apparatus based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the apparatus may include circuitry configured to perform an action.

[0011] Some example embodiments may be directed to a method. The method may include receiving, from a first network element, machine learning related resource availability that affects performance of a machine learning function, feature, or model performed by the first network element. The method may also include determining performance of the machine learning function, feature, or model based on the machine learning related resource availability that affects the performance of the machine learning function, feature, or model. The method may further include sending an indication indicating the performance of the machine learning functionality, feature, or model to a second network entity or the first network element.

[0012] Other exemplary embodiments may be directed to an apparatus. The apparatus may include at least one processor and at least one memory including computer program code. The at least one memory and the computer program code may be configured to, with the at least one processor, cause the apparatus to at least: receive, from a first network element, machine learning related resource availability that affects performance of a machine learning function, feature, or model performed by the first network element. The apparatus may also be caused to determine performance of the machine learning function, feature, or model based on the machine learning related resource availability that affects performance of the machine learning function, feature, or model. The apparatus may further be caused to send an indication indicating performance of the machine learning function, feature, or model to a second network entity or the first network element.

[0013] Other exemplary embodiments may be directed to an apparatus. The apparatus may include components for receiving, from a first network element, availability of machine learning related resources that affect performance of a machine learning function, feature, or model performed by the first network element. The apparatus may also include components for determining performance of the machine learning function, feature, or model based on the availability of machine learning related resources that affect performance of the machine learning function, feature, or model. The apparatus may further include components for sending an indication indicating performance of the machine learning function, feature, or model to a second network entity or the first network element.

[0014] According to other exemplary embodiments, a non-transitory computer readable medium may be encoded with instructions that, when executed in hardware, may perform a method. The method may include receiving, from a first network element, machine learning related resource availability that affects performance of a machine learning function, feature, or model performed by the first network element. The method may also include determining performance of the machine learning function, feature, or model based on the machine learning related resource availability that affects performance of the machine learning function, feature, or model. The method may further include sending an indication indicating the performance of the machine learning function, feature, or model to a second network entity or the first network element.

[0015] Other exemplary embodiments may be directed to a computer program product for performing a method. The method may include receiving, from a first network element, machine learning related resource availability that affects performance of a machine learning function, feature, or model performed by the first network element. The method may also include determining performance of the machine learning function, feature, or model based on the machine learning related resource availability that affects performance of the machine learning function, feature, or model. The method may further include sending an indication to a second network entity or the first network element indicating the performance of the machine learning function, feature, or model.

[0016] Other exemplary embodiments may be directed to an apparatus that may include circuitry configured to receive, from a first network element, availability of machine learning-related resources that affect performance of a machine learning function, feature, or model performed by the first network element. The apparatus may also include circuitry configured to determine performance of the machine learning function, feature, or model based on the availability of machine learning-related resources that affect performance of the machine learning function, feature, or model. The apparatus may further include circuitry configured to send an indication indicating performance of the machine learning function, feature, or model to a second network entity or the first network element.

[0017] Some example embodiments may be directed to a method. The method may include receiving an indication from a first network entity indicating performance of a machine learning function, feature, or model performed by a second network entity. The method may also include determining, based on the indication, an action to be performed by the second network element based on the indication. The method may further include configuring the second network element to perform an action at the second network element consistent with availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model.

[0018] Other exemplary embodiments may be directed to an apparatus. The apparatus may include at least one processor and at least one memory including computer program code. The at least one memory and the computer program code may be configured to, with the at least one processor, cause the apparatus to at least: receive an indication from a first network entity indicating performance of a machine learning function, feature, or model performed by a second network entity. The apparatus may also be caused to determine, based on the indication, an action to be performed by the second network element based on the indication. The apparatus may further be caused to configure the second network element to perform an action at the second network element consistent with availability of machine learning-related resources that affect performance of the machine learning function, feature, or model.

[0019] Other exemplary embodiments may be directed to an apparatus. The apparatus may include means for receiving an indication from a first network entity indicating performance of a machine learning function, feature, or model performed by a second network entity. The apparatus may also include means for determining, based on the indication, an action to be performed by the second network element based on the indication. The apparatus may further include means for configuring the second network element to perform an action at the second network element consistent with availability of machine learning-related resources that affect performance of the machine learning function, feature, or model.

[0020] According to other exemplary embodiments, a non-transitory computer readable medium may be encoded with instructions that, when executed in hardware, may perform a method. The method may include receiving an indication from a first network entity indicating performance of a machine learning function, feature, or model performed by a second network entity. The method may also include determining, based on the indication, an action to be performed by the second network element based on the indication. The method may further include configuring the second network element to perform an action at the second network element consistent with availability of machine learning-related resources that affect performance of the machine learning function, feature, or model.

[0021] Other exemplary embodiments may be directed to a computer program product for performing a method. The method may include receiving an indication from a first network entity indicating performance of a machine learning function, feature, or model performed by a second network entity. The method may also include determining, based on the indication, an action to be performed by the second network element based on the indication. The method may further include configuring the second network element to perform an action at the second network element consistent with availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model.

[0022] Other exemplary embodiments may be directed to an apparatus that may include circuitry configured to receive, from a first network entity, an indication of performance of a machine learning function, feature, or model performed by a second network entity. The apparatus may also include circuitry configured to determine, based on the indication, an action to be performed by the second network element based on the indication. The apparatus may further include circuitry configured to configure the second network element to perform an action at the second network element consistent with availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] For a proper understanding of the exemplary embodiments, reference should be made to the accompanying drawings, in which:

[0024] Figure 1 Example signal diagrams are shown in accordance with certain exemplary embodiments.

[0025] Figure 2 Another example signal diagram is shown in accordance with certain exemplary embodiments.

[0026] Figure 3 An example flow chart illustrating a method according to certain exemplary embodiments.

[0027] Figure 4 An example flow chart illustrating another method in accordance with certain exemplary embodiments.

[0028] Figure 5 An example flow chart illustrating yet another method in accordance with certain exemplary embodiments.

[0029] Figure 6 A set of apparatuses according to certain exemplary embodiments is shown. DETAILED DESCRIPTION

[0030] It will be readily appreciated that the components of certain exemplary embodiments as generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations. The following is a detailed description of some exemplary embodiments of systems, methods, apparatus, and computer program products for switching machine learning functions based on resource availability.

[0031] The features, structures, or characteristics of the exemplary embodiments described throughout this specification may be combined in any suitable manner in one or more exemplary embodiments. For example, the phrases "certain embodiments," "exemplary embodiments," "some embodiments," or other similar language used throughout this specification refer to the fact that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment. Therefore, the phrases "in certain embodiments," "exemplary embodiments," "in some embodiments," "in other embodiments," or other similar language appearing throughout this specification do not necessarily refer to the same set of embodiments, and the described features, structures, or characteristics may be combined in any suitable manner in one or more exemplary embodiments. Further, the terms "base station," "cell," "node," "gNB," "network," or other similar language throughout this specification may be used interchangeably. In addition, the terms machine learning (ML) and artificial intelligence (AI) may be used interchangeably throughout this specification.

[0032] As used herein, “at least one of: ” and “at least one of ” and similar expressions, if a list of two or more elements is connected by “and” or “or”, refer to at least any one of these elements, or at least any two or more of these elements, or at least all of these elements.

[0033] The technical specifications of the Third Generation Partnership Project (3GPP) describe the assessment of potential impacts on certain protocol aspects (e.g., Radio Access Network 2 (RAN2)), and consider aspects related to, for example, capability indication, configuration and control procedures (e.g., training / inference), and management of data and AI / ML models input per RAN1. Lifecycle management (LCM) of AI / ML models may include certain aspects such as, for example, data collection, model training, function / model identification, model transfer, model inference operations, function / model selection, activation, deactivation, handover, fallback operations, decisions of the network (NW) or user equipment (UE), function / model monitoring, model updates, and UE capabilities.

[0034] In the technical specifications of 3GPP, the LCM process can be based on an AI / ML model with a model ID, which has associated information and / or model functions for at least some AI / ML operations. In function-based LCM, the NW can indicate the activation, deactivation, fallback, or switching of the AI / ML function via 3GPP signaling (e.g., radio resource control (RRC), media access control-control element (MAC-CE) and / or downlink control information (DCI)). In some instances, the model may not be identified at the NW, and the UE may perform model-level LCM. For function identification, one or more functions may be defined in the AI / ML-enabled features, whereby the AI / ML-enabled features refer to the features in which AI / ML can be used.

[0035] For AI / ML function identification and function-based LCM of the UE part of the UE side model and / or the two-sided model, a function refers to an AI / ML-enabled feature / feature group (FG) enabled by a certain configuration or configurations. (One or more) configurations may be supported based on conditions indicated by UE capabilities, and the function-based LCM may operate based on at least one configuration of the AI / ML-enabled feature / FG or a specific configuration of the AI / ML-enabled feature / FG.

[0036] In model ID-based LCM, the model can be identified at the NW, and the NW / UE can activate, deactivate, select or switch a separate AI / ML model via the model ID. For AI / ML model identification and model ID-based LCM of the UE side model and / or the UE part of the bilateral model, the model ID-based LCM can operate based on the identified model. The model can be associated with a specific configuration or condition and additional conditions (e.g., scenarios, sites, and data sets) as determined / identified between the UE side and the NW side, where the specific configuration or condition is associated with the UE capability of the AI / ML-enabled feature / FG. From the RAN1 perspective, the AI / ML model identified by the model ID can be logical, and how it maps to (one or more) physical AI / ML models can depend on the implementation. The AI / ML model can be a data-driven algorithm that applies AI / ML techniques to generate a set of outputs based on a set of inputs.

[0037] According to certain exemplary embodiments, certain conditions may need to be defined to enable selection, activation, deactivation, switching, and fallback of practical and effective AI / ML solutions and AI / ML functions or models at the UE. UE capabilities associated with AI / ML-enabled features / FGs related to a certain AI / ML function and / or AI / ML model may be considered. Although the capabilities of the UE may be static (i.e., not expected to change within a relatively short time frame), there may also be dynamic conditions that affect whether the UE can utilize / switch to a certain AI / ML function / model at a certain time. For example, dynamic conditions may include, but are not limited to, the availability of resources, such as processing power to perform certain AI / ML functions at the UE.

[0038] In an example scenario, a certain AI / ML model of the UE may perform tasks such as, for example, location estimation as expected in a certain environment and situation. In this example, the UE may not be able to run the AI / ML model because the overloaded CPU is prioritized for other tasks at the UE. Therefore, the overloaded CPU may cause an unexpected but temporary drop in the performance of the AI / ML model. In this scenario, switching to another model of similar complexity may not help because the UE is already limited by its processing capabilities. Therefore, one option to solve this problem may be to switch to a model or fall back to a non-AI / ML solution with less computational complexity. However, there is currently no mechanism defined to address such model selection, activation, deactivation, switching and fallback on the UE side in such resource-constrained situations. Therefore, as described herein, certain exemplary embodiments may provide a way to select, activate, deactivate, switch and fallback (one or more) AI / ML-enabled features, functions or models on the UE side according to the conditions of the availability of AI / ML-related resources on the UE side. However, this solution described herein of certain exemplary embodiments may not be limited to the UE side, but may be applied to the network (e.g., gNB) side.

[0039] According to certain exemplary embodiments, a UE may be configured or triggered to monitor its AI / ML-related resources on the UE side that may affect the performance of an AI / ML-enabled feature, AI / ML function, or model, and indicate these dynamic conditions to an AI / ML performance monitoring entity (also referred to as a "performance monitoring entity") at the network (NW) (e.g., a core network entity (such as a network data analysis function (NWDAF)), a network node or function (such as, for example, a location management function (LMF) in the case of a positioning use case), or a gNB for a CSI or beam management use case). According to some exemplary embodiments, the performance monitoring entity may monitor the inference performance of the AI / ML model, and / or may derive monitoring metrics (e.g., statistics of (one or more) measurements compared to statistics associated with the training data; statistics of the difference between the model output and the provided ground truth labels). In certain exemplary embodiments, AI / ML-related resources may include, but are not limited to, for example, processing power, battery status, memory availability, etc.

[0040] In certain exemplary embodiments, once the performance monitoring entity receives a report of a dynamic condition from the UE, the performance monitoring entity may determine the AI / ML feature, function, or model performance based on the indicated AI / ML-related resource availability at the UE. The performance monitoring entity may also indicate the monitoring results to an AI / ML functional management entity (also referred to as a "functional management entity") at the NW. In some exemplary embodiments, the functional management entity may be another core network entity or function, such as, for example, a gNB, NWDAF, a functional management module (FMM) in the network.

[0041] According to certain exemplary embodiments, the function management entity may determine whether selection, activation, deactivation, switching or fallback of an AI / ML-enabled feature, AI / ML function or model should be performed at the UE based on the monitoring results. The function management entity may also provide relevant configurations to the UE, such as, for example, a configuration for switching to a certain AI / ML model that may be consistent with resource availability on the UE side. In this way, for certain exemplary embodiments, more effective decisions can be made on the LCM of AI / ML-enabled features by considering dynamic (and unforeseen) conditions related to AI / ML-enabled resources on the UE side.

[0042] Figure 1 An example signaling diagram is shown according to certain exemplary embodiments. Figure 1 and Figure 2As described in the examples of the present invention, the UE may be replaced by a radio access network (RAN) node (such as, for example, a gNB), and all described features of certain exemplary embodiments related to the UE may also be applicable to the gNB. In addition, similar to the UE, the gNB may also have (one or more) AI / ML-enabled features, (one or more) functions, or (one or more) models.

[0043] like Figure 1 As shown, at 115, the UE or RAN node 100 runs the AI / ML feature / model. At 120, the performance monitoring entity 105 (e.g., LMF in the case of the positioning use case) configures or triggers the UE or RAN node 100 to monitor its resource availability affecting the AI / ML feature / function / model on the UE or RAN node side.

[0044] In some exemplary embodiments, configuring and triggering the UE or RAN node 100 may be optional, and the UE or RAN node 100 may trigger itself to perform performance monitoring. In other exemplary embodiments, the performance monitoring entity may be a core network node / function, a RAN node / function, or a UE node / function.

[0045] At 125, the UE or RAN node 100 monitors its AI / ML related resources that may affect the performance of the AI / ML feature / model / function performance at the UE or RAN node side. In certain exemplary embodiments, the AI / ML related resources may include, but are not limited to, for example, processing power (e.g., at a graphics processing unit (GPU), a central processing unit (CPU), and / or a tensor processing unit (TPU)), available memory (e.g., random access memory (RAM) and / or read-only memory (ROM)), power (e.g., battery status), device input / output conditions (e.g., sensor blocking, sensor data loss, etc.), and / or device connection status (e.g., UL / DL throughput, link quality, etc.).

[0046] At 130, the UE or RAN node 100 may indicate the availability of AI / ML-related resources of the UE or RAN node 100 to the performance monitoring entity 105 via a binary indication, a duration of a (degraded) performance state, or detailed information / updates on the status of the AI / ML-related resources. In some exemplary embodiments, the indication may be related to a past performance state, a current state, or an expected future state. For example, in some exemplary embodiments, the binary indication may be an indication of whether there are sufficient AI / ML-related resources on the UE side to run the specified AI / ML-enabled features and / or AI / ML functions or models (e.g., indicated by function IDs, model IDs, etc.). In other exemplary embodiments, the duration of the (degraded) performance state may include, for example, how long the UE or RAN node is able or unable to perform the AI / ML function as expected. In some exemplary embodiments, the performance state may also be related to a space including, for example, a geographic area, a cell, etc. In addition, the performance state may be any performance state, including, but not limited to, for example, that resources are sufficient to perform a specific function. The performance of the UE or RAN node 100 may depend on its corresponding CPU scheduling. Additionally, in some exemplary embodiments, detailed information / updates regarding the status of AI / ML-related resources may include, for example, remaining battery power, CPU availability, memory usage, amount or percentage of power or memory (consumed / required or available), etc.

[0047] At 135, the performance monitoring entity 105 evaluates the information provided from the UE or RAN node 100 about the availability of AI / ML-related resources to determine the performance of the AI / ML-enabled features, AI / ML functions or models at the UE or RAN node 100. For example, in some exemplary embodiments, the performance may correspond to the ML inference performance. In an exemplary embodiment, if the ML task is to estimate the UE position ("direct AI / ML positioning"), it should be sufficiently accurate, and therefore, the performance metric may be the error of the absolute position estimate in a certain coordinate system. In addition, for the evaluation of AI / ML-assisted positioning in which the ML task is to estimate intermediate positioning-related features, various intermediate performance metrics may be used, including, for example, line of sight (LOS) classification accuracy, if the model output includes a hard-valued LOS / non-LOS indicator, where the LOS / non-LOS indicator is generated for a link between the UE and the total radiated power (TRP). If the model output includes a timing estimate (e.g., a time of arrival (ToA) reference signal time difference (RSTD)), the intermediate performance metric may also include a timing estimate accuracy (in meters). If the model includes angle estimation (e.g., angle of arrival (AoA) or angle of departure (AoD)), the intermediate performance metric may further include the angle estimation accuracy (in degrees). At 140, the performance monitoring entity 105 indicates the monitoring results to the function management entity 110 (e.g., NWDAF). In some exemplary embodiments, the function management entity 110 may be responsible for AI / ML feature / function / model switching / selection / (de)activation / fallback. In some exemplary embodiments, the function management entity 110 may be the same entity as the AI / ML performance monitoring entity.

[0048] At 145, the function management entity 110 may determine one or more of selection, activation, deactivation, switching, or fallback of an AI / ML feature / function / model at the UE or RAN node 100 based on the performance monitoring result. For example, at 150, the function management entity 110 may configure the UE or RAN node 100 to select, activate, deactivate, switch, or fallback to a certain AI / ML-enabled feature, AI / ML function or model, or non-AI / ML function. Figure 1 As shown, the functional management entity may send the configuration directly to the UE or RAN node 100, or first send the configuration to the performance monitoring entity 105, and the performance monitoring entity 105 sends the configuration to the UE or gNB node 100. Further, at 150, the functional management entity 110 may provide (one or more) timers or (one or more) thresholds for AI / ML related resources associated with the above configuration (e.g., to delay model switching when the CPU is available). In addition, at 155, the functional management entity 110 may optionally provide the UE or RAN node 100 with a new model that the UE or RAN node 100 should switch to. As shown in FIG. Figure 1 As shown, at 155, the new model may be sent directly to the UE or RAN node 100, or first provided to the performance monitoring entity 105, and the performance monitoring entity 105 sends the new model to the UE or RAN gNB 100.

[0049] Figure 2 Another example signal flow diagram according to certain exemplary embodiments is shown. For example, Figure 2 A multi-sided ML model is shown, wherein two or more entities involved in a two-sided (or multi-sided) ML model exchange information about whether their assigned ML-related tasks can be completed as planned or are experiencing some (temporal) conditions that affect their performance. According to certain exemplary embodiments, depending on the resource conditions at different sides of the model (e.g., at different UEs), the monitoring and / or functional management entity 210 may determine to distribute the load of (one or more) ML-related tasks across the multiple sides. For example, the monitoring and / or functional management entity 210 may allocate a less power-consuming part of the task to UEs suffering from low resource availability and allocate the remaining part of the task to other UEs with sufficient resources by configuring the relevant functions / models at different sides. Figure 2 The exemplary embodiments shown in Figure 1 Similar procedures to those shown in Figure 2 In the case of a multi-sided ML model, a message flows from one ML entity to another (e.g., one UE running a multi-sided ML model to another UE running the same multi-sided model, or to a network entity (gNB) running the same multi-sided model), indicating that the sharing of the ML model at one entity cannot be performed as planned.

[0050] At 215, UE1 200 runs the AI / ML features / models. At 220, UE1 200 determines whether it is possible to implement the sharing of the multilateral ML model with UE2 205. At 225, UE1 200 may directly indicate to UE2 205 the condition(s) that are degrading the ML computation sharing. Alternatively, at 225, UE1 200 may first indicate (i.e., send) to UE2 205 the condition(s) that are degrading the ML computation sharing, and UE2 205 may send the condition(s) that are degrading the ML computation sharing to the monitoring and / or function management entity 210. At 230, UE2 205 updates the AI / ML model sharing based on the indication from UE1 200. Similarly, at 235, the monitoring and / or function management entity 210 updates the ML model sharing based on the indication from UE1 200. At 240, the monitoring and / or function management entity 210 first indicates the ML model sharing update to UE2 205, and UE2 205 may send the ML model sharing update to UE1 200. Alternatively, the monitoring and / or function management entity 210 may send the ML model sharing update to UE1 200 directly.

[0051] Figure 3 An example flow chart of a method according to certain exemplary embodiments is shown. In an exemplary embodiment, Figure 3 The method may be performed by a network entity or a group of multiple network elements in a 3GPP system (such as LTE or 5G-NR). For example, in an exemplary embodiment, Figure 3 The method can be obtained by Figure 6 A similar UE or gNB is implemented in the apparatus 10 or 20 shown in FIG.

[0052] According to certain exemplary embodiments, Figure 3 The method may include: at 300, executing a machine learning function, feature, or model at a first network element. The method may also include: at 305, monitoring machine learning resources at the first network element that affect the performance of the machine learning function, feature, or model. In certain exemplary embodiments, the machine learning resources themselves may be required to perform / implement a certain ML function, feature, or model. The method may further include: at 310, based on the monitoring, indicating to a second network element the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. In addition, the method may include: at 315, receiving from a third network element a configuration of an action to be performed by the first network element based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the method may include performing an action at 320.

[0053] According to certain exemplary embodiments, the method may further include receiving a trigger from a second network element for monitoring the availability of resources of the first network element that affect the machine learning function, feature, or model. According to some exemplary embodiments, the machine learning resources may include at least one of processing power, available memory, power, device input or output condition, or device connection status. According to other exemplary embodiments, the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model may be provided by at least one of: a binary indication of whether there are sufficient machine learning-related resources at the first network element to run the specified machine learning-enabled feature, function, or model, a performance state of the machine learning function, a performance state related to time or space, or an update of the state of the machine learning-related resource, the update including the amount or percentage of power or memory consumed, required, or available.

[0054] In certain exemplary embodiments, the action to be performed by the first network element may include at least selecting, activating, deactivating, switching, or falling back to another machine learning function, feature, or model. In some exemplary embodiments, the selecting, activating, deactivating, switching, or falling back to another machine learning function may be performed based on at least one timer or at least one threshold value regarding the machine learning function, feature, or model. The action to be performed by the first network element may further or alternatively include at least receiving a new model to which the first network element should switch. In other exemplary embodiments, indicating the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model may include indicating the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model to another first network element that executes the same machine learning function, feature, or model. In further exemplary embodiments, the second network element and the third network element may be combined into a single network element.

[0055] Figure 4 An example flow chart of another method according to certain exemplary embodiments is shown. In an exemplary embodiment, Figure 4 The method may be performed by a network entity or a group of multiple network elements in a 3GPP system (such as LTE or 5G-NR). For example, in an exemplary embodiment, Figure 4 The method may be performed by a core network entity or function, such as, for example, Figure 6 A similar performance monitoring entity in the apparatus 10 or 20 shown in FIG.

[0056] According to certain exemplary embodiments, Figure 4The method may include: receiving, at 400, machine learning related resource availability that affects the performance of a machine learning function, feature, or model executed by the first network element from a first network element. The method may also include: determining, at 405, the performance of the machine learning function, feature, or model based on the machine learning related resource availability that affects the performance of the machine learning function, feature, or model. The method may further include: sending, at 410, an indication indicating the performance of the machine learning function, feature, or model to a second network entity or the first network element.

[0057] According to certain exemplary embodiments, the method may further include triggering the first network element to monitor resource availability of the first network element that affects the machine learning function, feature, or model. According to some exemplary embodiments, the availability of machine learning-related resources that affects the performance of the machine learning function, feature, or model may be received by at least one of: a binary indication of whether sufficient machine learning-related resources are available at the first network element to run the specified machine learning-enabled feature, function, or model, a performance status of the machine learning function, a performance status related to time or space, or an update of a status of a machine learning-related resource, the update including an amount or percentage of power or memory consumed, required, or available.

[0058] In some exemplary embodiments, the method may further include updating the machine learning function, feature, or model based on machine learning related resource availability that affects the performance of the machine learning function, feature, or model. In some exemplary embodiments, the method may also include the first network element notifying the updated machine learning function, feature, or model.

[0059] Figure 5 An exemplary flow chart of another method according to certain exemplary embodiments is shown. In the exemplary embodiment, Figure 5 The method may be performed by a network entity or a group of multiple network elements in a 3GPP system (such as LTE or 5G-NR). For example, in an exemplary embodiment, Figure 5 The method may be performed by a core network entity or function, such as, for example, Figure 6 A similar functional management entity in the apparatus 10 or 20 shown in FIG.

[0060] According to certain exemplary embodiments, Figure 5The method may include: receiving, at 500, an indication from a first network entity indicating performance of a machine learning function, feature, or model performed by a second network entity. The method may also include: determining, based on the indication, an action to be performed by the second network element based on the indication, at 505. The method may further include: configuring the second network element to perform an action consistent with availability of machine learning-related resources at the second network element that affect performance of the machine learning function, feature, or model, at 510.

[0061] According to certain exemplary embodiments, the action to be performed by the second network element comprises at least selecting, activating, deactivating, switching or falling back to another machine learning function, feature or model. According to certain exemplary embodiments, the selecting, activating, deactivating, switching or falling back to another machine learning function may be performed based on at least one timer or at least one threshold value for the machine learning function, feature or model. According to other exemplary embodiments, the action to be performed by the second network element may further or alternatively comprise at least receiving a new model to which the first network element should switch.

[0062] In some exemplary embodiments, the method may further include updating the machine learning function based on the availability of machine learning related resources that affect the performance of the machine learning function. In some exemplary embodiments, the method may also include notifying the second network element and the third network element of the updated machine learning function.

[0063] Figure 6 A set of apparatuses 10 and 20 according to certain exemplary embodiments are shown. In certain exemplary embodiments, the apparatuses 10 and 20 may be elements in a communication network or elements associated with such a network. For example, the apparatus 10 may be a UE or other similar radio communication computer device, and the apparatus 20 may be a network (i.e., a gNB, a performance network monitoring entity, or a functional management entity).

[0064] In some exemplary embodiments, the apparatuses 10 and 20 may include one or more processors, one or more computer-readable storage media (e.g., memory, storage device, etc.), one or more radio access components (e.g., modem, transceiver, etc.), and / or a user interface. In some exemplary embodiments, the apparatuses 10 and 20 may be configured to operate using one or more radio access technologies (such as GSM, LTE, LTE-A, NR, 5G, WLAN, WiFi, NB-IoT, Bluetooth, NFC, MulteFire, and / or any other radio access technology). It should be noted that a person of ordinary skill in the art will understand that the apparatuses 10 and 20 may include one or more processors, one or more computer-readable storage media (e.g., memory, storage device, etc.), one or more radio access components (e.g., modem, transceiver, etc.), and / or a user interface. Figure 6 Components or features not shown.

[0065] As in Figure 6 As shown in the example of , the devices 10 and 20 may include or be coupled to processors 12 and 22 for processing information and executing instructions or operations. The processors 12 and 22 may be any type of general or special purpose processor. In fact, as examples, the processors 12 and 22 may include one or more of a general purpose computer, a special purpose computer, a microprocessor, a DSP, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), and a processor based on a multi-core processor architecture. Although in Figure 6 1 and 22, but multiple processors may be utilized according to other exemplary embodiments. For example, it should be understood that in some exemplary embodiments, the devices 10 and 20 may include two or more processors, which may form a multi-processor system that can support multi-processing (e.g., in this case, the processor 12 may represent a multi-processor). According to some exemplary embodiments, the multi-processor system may be tightly coupled or loosely coupled (e.g., to form a computer cluster).

[0066] Processors 12 and 22 may perform functions associated with the operation of devices 10 and 20, including, as some examples, precoding of antenna gain / phase parameters, encoding and decoding of individual bits forming communication messages, formatting of information, and overall control of devices 10 and 20, including in Figures 1 to 5 The process and examples shown in .

[0067] The devices 10 and 20 may further include or be coupled to memories 14 and 24 (internal or external), which may be coupled to the processors 12 and 24, respectively, for storing information and instructions that can be executed by the processors 12 and 24. The memories 14 and 24 may be one or more memories and any type suitable for the local application environment, and may be implemented using any suitable volatile or non-volatile data storage technology, such as semiconductor-based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and / or removable memory. For example, the memories 14 and 24 may include any combination of random access memory (RAM), read-only memory (ROM), static storage (such as a magnetic disk or optical disk), a hard disk drive (HDD), or any other type of non-transitory machine or computer readable medium. The instructions stored in the memories 14 and 24 may include program instructions or computer program code, which, when executed by the processors 12 and 22, enables the devices 10 and 20 to perform tasks as described herein.

[0068] In certain exemplary embodiments, the devices 10 and 20 may further include or be coupled to a (internal or external) drive or port configured to accept and read an external computer-readable storage medium, such as an optical disk, a USB drive, a flash drive, or any other storage medium. For example, the external computer-readable storage medium may store programs for execution by the processors 12 and 22 and / or the devices 10 and 20 to perform Figures 1 to 5 Computer programs or software of any methods and examples shown.

[0069] In some exemplary embodiments, the devices 10 and 20 may also include or be coupled to one or more antennas 15 and 25 for receiving downlink signals and transmitting from the devices 10 and 20 via the UL. The devices 10 and 20 may further include transceivers 18 and 28 configured to send and receive information. The transceivers 18 and 28 may also include a radio interface (e.g., a modem) coupled to the antennas 15 and 25. The radio interface may correspond to a plurality of radio access technologies, including one or more of GSM, LTE, LTE-A, 5G, NR, WLAN, NB-IoT, Bluetooth, BT-LE, NFC, RFID, UWB, etc. The radio interface may include other components, such as filters, converters (e.g., digital-to-analog converters, etc.), symbol demappers, signal shaping components, inverse fast Fourier transform (IFFT) modules, etc., to process symbols carried by the downlink or UL, such as OFDMA symbols.

[0070] For example, the transceivers 18 and 28 may be configured to modulate information onto a carrier waveform for transmission by the antennas 15 and 25, and to demodulate information received via the antennas 15 and 25 for further processing by other elements of the devices 10 and 20. In other exemplary embodiments, the transceivers 18 and 28 may be capable of directly sending and receiving signals or data. Additionally or alternatively, in some exemplary embodiments, the device 10 may include input and / or output devices (I / O devices). In some exemplary embodiments, the devices 10 and 20 may also include a user interface, such as a graphical user interface or a touch screen.

[0071] In certain exemplary embodiments, the memory 14 and 34 store software modules that provide functionality when executed by the processors 12 and 22. The modules may include, for example, an operating system that provides operating system functionality to the devices 10 and 20. The memory may also store one or more functional modules, such as applications or programs, to provide additional functionality to the devices 10 and 20. The components of the devices 10 and 20 may be implemented in hardware or as any suitable combination of hardware and software. According to certain exemplary embodiments, the devices 10 and 20 may optionally be configured to communicate with each other (in any combination) via a wireless or wired communication link 70 according to any radio access technology (e.g., NR).

[0072] According to some exemplary embodiments, the processors 12 and 22 and the memories 14 and 24 may be included in a processing circuit or a control circuit or may form a part of a processing circuit or a control circuit. In addition, in some exemplary embodiments, the transceivers 18 and 28 may be included in a transceiver circuit or may form a part of a transceiver circuit.

[0073] For example, in certain exemplary embodiments, the device 10 may be controlled by the memory 14 and the processor 12 to perform a machine learning function, feature, or model at the device. The device 10 may also be controlled by the memory 14 and the processor 12 to monitor machine learning resources at the device that affect the performance of the machine learning function, feature, or model. The device 10 may be further controlled by the memory 14 and the processor 12 to indicate to a first network element, based on monitoring, the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. In addition, the device 10 may be controlled by the memory 14 and the processor 12 to receive from a second network element a configuration of an action to be performed by the device based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the device 10 may be controlled by the memory 14 and the processor 12 to perform an action.

[0074] In certain other exemplary embodiments, the apparatus 20 may be controlled by the memory 24 and the processor 22 to receive from the first network element the availability of machine learning related resources that affects the performance of the machine learning function, feature, or model performed by the first network element. The apparatus 20 may also be controlled by the memory 24 and the processor 22 to determine the performance of the machine learning function, feature, or model based on the availability of machine learning related resources that affects the performance of the machine learning functionality, feature, or model. The apparatus 20 may further be controlled by the memory 24 and the processor 22 to send an indication indicating the performance of the machine learning function, feature, or model to the second network entity or the first network element.

[0075] In a further exemplary embodiment, the apparatus 20 may be controlled by the memory 24 and the processor 22 to receive an indication from the first network entity indicating the performance of a machine learning function, feature, or model performed by the second network entity. The apparatus 20 may also be controlled by the memory 24 and the processor 22 to determine, based on the indication, an action to be performed by the second network element based on the indication. The apparatus 20 may further be controlled by the memory 24 and the processor 22 to configure the second network element to perform an action consistent with the availability of machine learning related resources at the second network element that affect the performance of the machine learning function, feature, or model.

[0076] In some exemplary embodiments, an apparatus (e.g., apparatus 10 and / or apparatus 20) may include components for performing the methods, processes, or any variations discussed herein. Examples of components may include one or more processors, memories, controllers, transmitters, receivers, and / or computer program code for causing the execution of operations.

[0077] Certain exemplary embodiments may be directed to an apparatus comprising components for performing any of the methods described herein, including, for example, components for executing a machine learning function, feature, or model at the apparatus. The apparatus may also include components for monitoring machine learning resources at the apparatus that affect the performance of the machine learning function, feature, or model. The apparatus may further include components for indicating to a first network element, based on the monitoring, the availability of machine learning-related resources that affect the performance of the machine learning functionality, feature, or model. In addition, the apparatus may include: components for receiving, from a second network element, a configuration of an action to be performed by the apparatus based on the availability of machine learning-related resources that affect the performance of the machine learning function, feature, or model. Further, the apparatus may include components for performing an action.

[0078] Additional exemplary embodiments may provide an apparatus including means for receiving, from a first network element, availability of machine learning related resources that affect performance of a machine learning function, feature, or model performed by the first network element. The apparatus may also include means for determining performance of the machine learning function, feature, or model based on the availability of machine learning related resources that affect performance of the machine learning function, feature, or model. The apparatus may further include means for sending an indication indicating performance of the machine learning function, feature, or model to a second network entity or the first network element.

[0079] Further exemplary embodiments may be directed to an apparatus that includes means for receiving, from a first network entity, an indication of performance of a machine learning function, feature, or model performed by a second network entity. The apparatus may also include means for determining, based on the indication, an action to be performed by the second network element based on the indication. The apparatus may further include means for configuring the second network element to perform an action consistent with availability of machine learning-related resources at the second network element that affect performance of the machine learning function, feature, or model.

[0080] Certain exemplary embodiments described herein provide several technical improvements, enhancements and / or advantages. For example, in some exemplary embodiments, a method for selecting, activating, deactivating, switching and / or falling back on AI / ML-enabled features, functions or models at a UE or RAN node based on conditions of AI / ML-related resource availability at the UE or RAN node side may be provided. According to other exemplary embodiments, by considering resource availability conditions at the ML execution entity, the performance of ML-enabled features (e.g., better accuracy of UE location estimation, etc.) may be improved by timely configuration of ML functions.

[0081] The computer program product may include one or more computer executable components that are configured to perform some exemplary embodiments when the program is run. The one or more computer executable components may be at least one software code or part thereof. Modifications and configurations required to implement the functions of certain exemplary embodiments may be performed as (one or more) routines that may be implemented as added or updated (one or more) software routines. (One or more) software routines may be downloaded to a device.

[0082] As an example, the software or computer program code or part thereof may be in source code form, object code form or some intermediate form, and may be stored in some carrier, distribution medium or computer readable medium, which may be any entity or device capable of carrying the program. For example, such carriers may include recording media, computer memory, read-only memory, optoelectronic and / or electrical carrier signals, telecommunication signals, and software distribution packages. Depending on the processing power required, the computer program may be executed in a single electronic digital computer, or it may be distributed among multiple computers. The computer readable medium or computer readable storage medium may be a non-transient medium.

[0083] In other exemplary embodiments, the functions may be performed by hardware or circuitry included in a device (e.g., device 10 or device 20), such as by using an application specific integrated circuit (ASIC), a programmable gate array (PGA), a field programmable gate array (FPGA), or any other combination of hardware and software. In yet another exemplary embodiment, the functions may be implemented as signals, non-tangible components that may be carried by electromagnetic signals downloaded from the Internet or other networks.

[0084] According to certain exemplary embodiments, an apparatus (such as a node, a device, or a corresponding component) may be configured as a circuit, a computer or a microprocessor (such as a single-chip computer element), or a chipset, which includes at least a memory for providing storage capacity for arithmetic operations and an operation processor for performing arithmetic operations.

[0085] Those of ordinary skill in the art will readily appreciate that the present disclosure as discussed above may be practiced with procedures in a different order and / or with hardware elements in configurations different from those disclosed. Therefore, although the present disclosure has been described based on these exemplary embodiments, it will be apparent to those skilled in the art that certain modifications, variations, and alternative constructions will be apparent while remaining within the spirit and scope of the exemplary embodiments. Although the above embodiments relate to 5G NR and LTE technologies, the above embodiments may also be applied to any other current or future 3GPP technologies, such as Advanced LTE and / or fourth generation (4G) technologies.

[0086] Partial glossary:

[0087] 3GPP Third Generation Partnership Project

[0088] 5G 5th Generation

[0089] 5GCN 5G Core Network

[0090] 5GS 5G System

[0091] AI

[0092] BS Base Station

[0093] CPU Central Processing Unit

[0094] CSI Channel State Information

[0095] DL Downlink

[0096] eNB Enhanced Node B

[0097] E-UTRAN Evolved UTRAN

[0098] FG Feature Group

[0099] gNB 5G or Next Generation Node B

[0100] GPU Graphics Processing Unit

[0101] ID

[0102] KPI Key Performance Indicator

[0103] LCM Lifecycle Management

[0104] LMF Location Management Function

[0105] LTE Long Term Evolution

[0106] ML Machine Learning

[0107] NR New Radio

[0108] NW Network

[0109] RAM Random Access Memory

[0110] RAN Radio Access Network

[0111] ROM Read Only Memory

[0112] TPU Tensor Processing Unit

[0113] UE User Equipment

[0114] UL Uplink

Claims

1. A method for communication, comprising: executing a machine learning function, feature, or model at the first network element; monitoring machine learning resources at the first network element that affect performance of the machine learning function, feature, or model; Based on the monitoring, indicating to a second network element availability of machine learning-related resources that impact performance of the machine learning function, feature, or model; receiving, from a third network element, a configuration of actions to be performed by the first network element based on availability of the machine learning-related resources that affect performance of the machine learning function, feature, or model; and Perform the action.

2. The method according to claim 1, further comprising: A trigger is received from the second network element for monitoring resource availability of the first network element that affects the machine learning function, feature, or model.

3. The method according to claim 1, wherein: The machine learning resource includes at least one of the following: Processing power, Availability memory, electricity, Equipment input or output conditions, or Device connection status.

4. The method according to claim 1, wherein: The availability of the machine learning related resources that affect the performance of the machine learning function, feature, or model is provided by at least one of the following: a binary indication of whether sufficient machine learning related resources exist at the first network element to run the specified machine learning enabled feature, function or model, a performance state of the machine learning function, the performance state being related to time or space, or Updates to the status of the machine learning-related resources, including the amount or percentage of power or memory consumed, required, or available.

5. The method according to any one of claims 1 to 4, wherein: The action to be performed by the first network element includes at least one of the following: selecting, activating, deactivating, switching, or falling back to another machine learning function, feature, or model, wherein the selecting, activating, deactivating, switching, or falling back to the another machine learning function is performed based on at least one timer or at least one threshold value related to the machine learning function, feature, or model, or A new model is received to which the first network element should switch.

6. The method according to any one of claims 1 to 4, wherein: Indicating the availability of the machine learning-related resources that affect the performance of the machine learning function, feature, or model includes: indicating the availability of the machine learning-related resources that affect the performance of the machine learning function, feature, or model to another first network element that executes the same machine learning function, feature, or model.

7. The method according to any one of claims 1 to 4, wherein: The second network element and the third network element are combined into a single network element.

8. A method for communication, comprising: receiving, from a first network element, machine learning-related resource availability that affects performance of a machine learning function, feature, or model executed by the first network element; determining the performance of the machine learning function, feature, or model based on the availability of the machine learning-related resources that affect the performance of the machine learning function, feature, or model; and An indication is sent to a second network entity or the first network element indicating a performance of the machine learning function, feature, or model.

9. The method according to claim 8, further comprising: The first network element is triggered to monitor resource availability of the first network element that affects the machine learning function, feature, or model.

10. The method according to claim 8, wherein: The availability of the machine learning related resources that affect the performance of the machine learning function, feature, or model is received via at least one of: a binary indication of whether sufficient machine learning related resources exist at the first network element to run the specified machine learning enabled feature, function or model, a performance state of the machine learning function, the performance state being related to time or space, or Updates to the status of the machine learning-related resources, including the amount or percentage of power or memory consumed, required, or available.

11. The method according to any one of claims 8 to 10, further comprising: Updating the machine learning function, feature, or model based on the availability of machine learning related resources that affect the performance of the machine learning function, feature, or model.

12. The method according to claim 11, further comprising: The first network element is notified of the updated machine learning function, feature, or model.

13. A method for communication, comprising: receiving, from the first network entity, an indication indicating performance of a machine learning function, feature, or model performed by the second network entity; determining, based on the indication, an action to be performed by the second network element based on the indication; and The second network element is configured to perform actions consistent with availability of machine learning-related resources at the second network element that affect performance of the machine learning function, feature, or model.

14. The method according to claim 13, wherein: The actions to be performed by the second network element include at least one of the following: selecting, activating, deactivating, switching, or falling back to another machine learning function, feature, or model, wherein the selecting, activating, deactivating, switching, or falling back to the another machine learning function is performed based on at least one timer or at least one threshold value related to the machine learning function, feature, or model, or A new model is received to which the first network element should switch.

15. The method according to claim 13 or 14, further comprising: The machine learning function is updated based on the availability of machine learning related resources that affect the performance of the machine learning function.

16. The method according to claim 15, further comprising: The second network element and the third network element are notified of the updated machine learning functionality.

17. An apparatus for communication, comprising: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, together with the at least one processor, cause the apparatus to at least: executing a machine learning function, feature, or model at the device; monitoring machine learning-related resources at the device that affect performance of the machine learning function, feature, or model; Based on the monitoring, indicating to the first network element availability of machine learning-related resources that impact performance of the machine learning function, feature, or model; receiving, from a second network element, a configuration of an action to be performed by the device based on availability of the machine learning-related resources that affects performance of the machine learning function, feature, or model; and Perform the action.

18. The device according to claim 17, wherein: The at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus to at least: A trigger is received from the first network element for monitoring resource availability of the device affecting the machine learning function, feature, or model.

19. The device according to claim 17, wherein: The machine learning resource includes at least one of the following: Processing power, Availability memory, electricity, Equipment input or output conditions, or Device connection status.

20. The device according to claim 17, wherein: The availability of the machine learning related resources that affect the performance of the machine learning function, feature, or model is provided by at least one of the following: a binary indication of whether sufficient machine learning related resources exist at the device to run the specified machine learning enabled feature, function, or model, a performance state of the machine learning function, the performance state being related to time or space, or Updates to the status of the machine learning-related resources, including the amount or percentage of power or memory consumed, required, or available.

21. The device according to any one of claims 17 to 20, wherein: The action to be performed by the device includes at least one of the following: selecting, activating, deactivating, switching, or falling back to another machine learning function, feature, or model, wherein the selecting, activating, deactivating, switching, or falling back to the another machine learning function is performed based on at least one timer or at least one threshold value related to the machine learning function, feature, or model, or Receive a new model to which the device should switch.

22. The device according to any one of claims 17 to 20, wherein: Indicating the availability of the machine learning-related resources that affect the performance of the machine learning function, feature, or model includes: indicating the availability of the machine learning-related resources that affect the performance of the machine learning function, feature, or model to another device that executes the same machine learning function, feature, or model.

23. The device according to any one of claims 17 to 20, wherein: The first network element and the second network element are combined into a single network element.

24. An apparatus for communication, comprising: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, together with the at least one processor, cause the apparatus to at least: receiving, from a first network element, machine learning-related resource availability that affects performance of a machine learning function, feature, or model executed by the first network element; determining the performance of the machine learning function, feature, or model based on the availability of the machine learning-related resources that affect the performance of the machine learning function, feature, or model; and An indication is sent to a second network entity or the first network element indicating a performance of the machine learning function, feature, or model.

25. The device according to claim 24, wherein: The at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus to at least: The first network element is triggered to monitor resource availability of the first network element that affects the machine learning function, feature, or model.

26. The device according to claim 24, wherein: The availability of the machine learning related resources that affect the performance of the machine learning function, feature, or model is received via at least one of: a binary indication of whether sufficient machine learning related resources exist at the first network element to run the specified machine learning enabled feature, function or model, a performance state of the machine learning function, the performance state being related to time or space, or Updates to the status of the machine learning-related resources, including the amount or percentage of power or memory consumed, required, or available.

27. The device according to any one of claims 24 to 26, wherein: The at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus to at least: Updating the machine learning function, feature, or model based on the availability of machine learning related resources that affect the performance of the machine learning function, feature, or model.

28. The device according to claim 27, wherein The at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus to at least: The first network element is notified of the updated machine learning function, feature, or model.

29. An apparatus for communication, comprising: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, together with the at least one processor, cause the apparatus to at least: receiving, from the first network entity, an indication indicating performance of a machine learning function, feature, or model performed by the second network entity; determining, based on the indication, an action to be performed by the second network element based on the indication; and The second network element is configured to perform actions consistent with availability of machine learning-related resources at the second network element that affect performance of the machine learning function, feature, or model.

30. The device according to claim 29, wherein The actions to be performed by the second network element include at least one of the following: selecting, activating, deactivating, switching, or falling back to another machine learning function, feature, or model, wherein the selecting, activating, deactivating, switching, or falling back to the another machine learning function is performed based on at least one timer or at least one threshold value related to the machine learning function, feature, or model, or A new model is received to which the first network element should switch.

31. The device according to claim 29 or 30, wherein: The at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus to at least: The machine learning function is updated based on the availability of machine learning related resources that affect the performance of the machine learning function.

32. The device according to claim 31, wherein The at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus to at least: The second network element and the third network element are notified of the updated machine learning functionality.

Images