Transfer learning in application layer machine learning tasks in wireless communication networks

Abstract

Various aspects of this disclosure relate to a method (900) performed by a first network entity, comprising: receiving (902) a first request from a second network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; discovering (904) a third network entity for indicating the availability of the pre-trained ML model based on the one or more criteria; obtaining (906) first information on one or more available pre-trained ML models from the third network entity based on the one or more criteria; and providing (908) second information to the second network entity, the second information indicating the one or more available pre-trained ML models as pre-trained ML models for the transfer learning for the ML model task.

Description

Technical Field

[0001] The subject matter disclosed herein generally relates to the field of implementing transfer learning in application-layer machine learning tasks within wireless communication networks. This invention defines a first network entity, a second network entity, and a third network entity for wireless communication. The invention further defines a method performed by the first network entity, a method performed by the second network entity, and a method performed by the third network entity. Background Technology

[0002] A wireless communication system may include one or more network communication devices, such as base stations, which can support wireless communication with one or more user communication devices (which may also be referred to as user equipment (UE) or other suitable terms). The wireless communication system can support wireless communication with one or more user communication devices by utilizing the resources of the wireless communication system (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers, etc.)). Furthermore, the wireless communication system can support wireless communication across various radio access technologies, including third-generation (3G) radio access technology, fourth-generation (4G) radio access technology, fifth-generation (5G) radio access technology, and other suitable radio access technologies beyond 5G (e.g., sixth-generation (6G)).

[0003] Transfer learning (TL) is a technique in machine learning where a model trained on one task is used as a starting point for a model on a second task. This is useful when the second task is similar to the first task or when limited data is available for the second task. By using features learned from the first task as a starting point, the model can learn faster and more efficiently on the second task. This also helps prevent overfitting, as the model has learned general features that might be useful in the second task. Summary of the Invention

[0004] The article “a” preceding an element is not a limitation, but should be understood to refer to “at least one” of these elements or “one or more” of these elements. The terms “a,” “at least one,” “one or more,” and “at least one of one or more” are interchangeable. As used herein (included in the claims), the word “or” used in a list of items (e.g., a list of items beginning with phrases such as “at least one of…”, “one or more of…”, or “one or both of…”) indicates an inclusive list, such that (e.g.) a list of at least one of A, B, or C means A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Furthermore, as used herein, the phrase “based on” should not be construed as referring to a closed set of conditions. For example, without departing from the scope of this disclosure, an exemplary step described as “based on condition A” may be based on both condition A and condition B. In other words, as used herein, the phrase “based on” should be interpreted in the same manner as the phrase “at least partially based on.” Additionally, as used herein (included in the claims), “group” may include one or more elements.

[0005] A first network entity for wireless communication is provided, comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the first network entity: receives from a second network entity a first request to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; based on the one or more criteria, discovers a third network entity for indicating the availability of the pre-trained ML model; based on the one or more criteria, obtains first information about one or more available pre-trained ML models from the third network entity; and provides second information to the second network entity, the second information indicating the one or more available pre-trained ML models as pre-trained ML models for the transfer learning for the ML model task.

[0006] A second network entity for wireless communication is further provided, comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the second network entity: sends a first request to a first network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; and receives second information from the first network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for the transfer learning of the ML model task.

[0007] A third network entity for wireless communication is further provided, comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the third network entity: receives from a first network entity a second request for an available pre-trained ML model, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; determines first information of one or more available pre-trained ML models based on the one or more criteria; and sends a second response including the first information of the one or more available pre-trained ML models to the first network entity.

[0008] A method further provided by a first network entity includes: receiving from a second network entity a first request to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; based on the one or more criteria, discovering a third network entity for indicating the availability of the pre-trained ML model; based on the one or more criteria, obtaining first information about one or more available pre-trained ML models from the third network entity; and providing second information to the second network entity, the second information indicating the one or more available pre-trained ML models as pre-trained ML models for the transfer learning for the ML model task.

[0009] A method performed by a second network entity is further provided, comprising: sending a first request to a first network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; and receiving second information from the first network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for the transfer learning of the ML model task.

[0010] A further method is provided, performed by a third network entity, comprising: receiving a second request for an available pre-trained ML model from a first network entity, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; determining first information of one or more available pre-trained ML models based on the one or more criteria; and sending a second response including the first information of the one or more available pre-trained ML models to the first network entity. Attached Figure Description

[0011] Figure 1 Examples of wireless communication systems according to aspects of this disclosure are described.

[0012] Figure 2Examples of advanced transfer learning operations based on aspects of this disclosure are provided.

[0013] Figure 3 This describes an example of a high-level architecture based on aspects of this disclosure.

[0014] Figure 4 This describes an instance of a message stream enabled by TL triggered by ADAE according to aspects of this disclosure.

[0015] Figure 5 This describes an instance of a message stream enabled by a TL triggered by a VAL according to aspects of this disclosure.

[0016] Figure 6 An example of a user equipment (UE) 600 according to aspects of this disclosure is described.

[0017] Figure 7 An example of processor 700 according to aspects of this disclosure is described.

[0018] Figure 8 Examples of network equipment (NE) 800 according to aspects of this disclosure are described.

[0019] Figure 9 A flowchart illustrating the method performed by NE according to aspects of this disclosure.

[0020] Figure 10 A flowchart illustrating the method performed by NE according to aspects of this disclosure.

[0021] Figure 11 A flowchart illustrating the method performed by NE according to aspects of this disclosure. Detailed Implementation

[0022] The 3rd Generation Partnership Project (3GPP) has discussed the use of artificial intelligence (AI) as a tool to optimize existing features and procedures by making them smarter, or to assist AI applications in communicating via mobile communication systems, in various areas (RAN, core, service enablement layer).

[0023] 3GPP's work at the core and application layers has been implemented in 3GPP SA2, which began researching AI / ML support with Rel-16. This focuses on two main aspects: 1) AI / ML support in Network Data Analysis Function (NWDAF); and 2) Network assistance for AI / ML services.

[0024] Regarding the first key aspect, network analytics and AI / ML are deployed in the 5G core network via the introduction of NWDAF. NWDAF is designed to support various analytics types that can be distinguished using different analytics IDs (e.g., “UE Mobility,” “NF Load,” etc.), as described in 3GPP Technical Specification (TS) 23.288, entitled “Architecture enhancements for 5G System to support network data analytics services.” Each NWDAF can support one or more analytics IDs and can have the following functions: (i) AI / ML inference called NWDAF AnLF, or (ii) AI / ML training called NWDAF MTLF, or (iii) both. For the Federated Learning (FL) use case, in 3GPP Release 18, 3GPP defines federated learning in different NWDAF MTLFs, where ML model training runs in multiple local MTLFs.

[0025] Regarding the second main aspect, as part of AIMLSys, enhancements to 5GC (as specified in Clause 5.46 of 3GPP TS 23.501 entitled "System Architecture for the 5G system") are designated to assist AI / ML operations in the application layer (between one or more AI / ML users and AI / ML servers). In this direction, Network Open Functions (NEF) can assist AI / ML application servers in scheduling available UEs to participate in AI / ML operations (e.g., federated learning). Furthermore, 5GC can assist in selecting UEs as FL clients by providing a list of target member UEs and then subscribing to NEF to be notified of a subset of UEs meeting specific screening criteria (i.e., a candidate UE list).

[0026] 3GPP SA4 continues its research on AI and ML in media (see 3GPP Technical Report (TR) 26.927, titled "Study on Artificial Intelligence and Machine Learning in 5G mediaservices"). This study aims to identify relevant interoperability requirements and implementation constraints for AI / ML in 5G media services. This research includes media-based AI / ML use cases and architectural considerations related to media services.

[0027] 3GPP SA5 in Rel-17 also provides work on AI / ML management (see 3GPP TS 28.105, titled "Management and orchestration; Artificial Intelligence / Machine Learning management"), which focuses on AI / ML capabilities (e.g., ML training MnS / MF) on the Operations, Management, and Maintenance (OAM) side. 3GPP TS 28.105 specifies AI / ML management capabilities and services for 5GS in which AI / ML is used, including management and orchestration (e.g., MDA, see 3GPP TS28.104, titled "Management and orchestration; management data analytics (MDA)"). In this work, the ML Training Function (MLT) is introduced, acting as a producer of ML training MnS, which consumes various data from the 5GS for ML training purposes and provides the training output to other management functions in OAM.

[0028] 3GPP SA6 in Rel-19 continues the research on AIML enablers (which can be logically placed in ADAES or the new SEAL server) to support AI / ML services via the service enablement layer. This research is discussed in 3GPP TR 23.700-82, entitled "Study on Application Layer support for AI / ML services".

[0029] In 3GPP TR 23.700-82 (AIMLAPP study in Rel-19 of 3GPP SA6), key issues concerning supporting transfer learning in 3GPP systems are raised. A key question that arises is how to support transfer learning at the application enablement layer.

[0030] In 3GPP TS 22.261, titled "Service requirements for the 5G system", clause 6.40 states that the 5G system needs to support at least three types of AI / ML operations: AI / ML operation decomposition between AI / ML endpoints; distributed / federated learning on the 5G system; and AI / ML model / data distribution and sharing on the 5G system.

[0031] Regarding the distribution and sharing of AI / ML models / data on 5G systems, it is expected that AI / ML consumers will be able to use ML models (a set of candidate models available from NW endpoints) to better adapt to changes in tasks and environments.

[0032] However, in real-world deployments involving AI / ML operations, the number of available models is insufficient to cover all possible environmental conditions within the network. Therefore, using models out of the box implies a lack of accuracy / precision during the inference phase unless a fine-tuning phase is provided to ensure the baseline model adapts quickly to current conditions using a smaller amount of data (compared to the amount of data used to train the baseline model). In this regard, transfer learning (TL) is considered a technique to address the lack of labeled data for training models, which is the first limitation faced when (re)training models. In TL, the model is trained (to solve a specific task) using information from a "similar" domain (the so-called source domain) where sufficient information is available. Through TL, parts of the model trained in the source domain are fine-tuned with the scarce information available in the target domain. In this way, the model reuses its understanding of the problem's structure and does not have to learn all the low-level features / structures of the problem from scratch: it will only need to learn higher-level structures / relationships that typically require less labeled data. TL assumes that the baseline model is already available (in the NW endpoint) and can be fine-tuned to quickly perform the same or similar tasks (e.g., prediction, classification, etc.) in the target domain with less training data.

[0033] First, consider that: In 3GPP TS 23.288, Clauses 6.2B.5, 6.2B.6, and 6.2B.7, some operations that allow storing ML models to the Analytics Data Repository Function (ADRF) / retrieving ML models from the ADRF can be consumed by network functions containing examples of Model Training Logic Functions (MTLF); and according to Clause 5.3 of 3GPP TS 23.436 (titled "Functional architecture and information flows for application data analytics enablement service"), the internal architecture of the ADAE (Application Data Analytics Enabler) can contain examples of ADRFs, the so-called A-ADRF (Application Layer-Analysis Data Repository Function), which can enable transfer learning scenarios between and within VAL domains.

[0034] As mentioned above, transfer learning is used to support entities in simplifying the training process (i.e., if it is not possible or if there is insufficient training data and time). When applying transfer learning (TL) to features in 5G and later systems, different use cases can be assumed for the affected operations.

[0035] The first use case is to employ TL in AI / ML enabled analytics between network analytics functions in the core (between NWDAF / MTLF).

[0036] A further use case is to employ TL in AI / ML enabled analytics between SEAL ADAES (such as SA6, ADAES defined in 3GPP TS 23.436).

[0037] A further use case is to employ TL between AF and NF, such as in cross-domain analysis of AI-assisted operations.

[0038] A further use case is to use TL to assist AI / ML operations between VAL servers or Af.

[0039] A further use case is to use TL to assist AI / ML operations between the AF / VAL server and VAL UE or between VAL UEs.

[0040] For TL enablement in 3GPP systems that focus more on the application layer, the following questions can be identified: how to identify whether a pre-trained model can be used for another ML model task (i.e., a new ML job) and how to ensure that this solution will be beneficial; and how to configure the TL enablement entity for the distribution of pre-trained ML models.

[0041] This disclosure describes the identification of candidate pre-trained ML models and their criterion-based selection. The configuration of the entities involved in the distribution of pre-trained ML models (i.e., the source, destination, coordinator / registry of the pre-trained ML models) is further described herein. Licensing and billing aspects are also included. Such solutions enable the benefits of transfer learning as discussed above within 3GPP systems.

[0042] This disclosure is described in the context of wireless communication systems.

[0043] Figure 1This describes an example of a wireless communication system 100 according to aspects of this disclosure. The wireless communication system 100 may include one or more NEs 102, one or more UEs 104, and a core network (CN) 106. The wireless communication system 100 may support various radio access technologies. In some embodiments, the wireless communication system 100 may be a 4G network, such as an LTE network or an LTE-A network. In some other embodiments, the wireless communication system 100 may be an NR network, such as a 5G network, a 5G-A network, or a 5G Ultra Wideband (5G-UWB) network. In other embodiments, the wireless communication system 100 may be a combination of 4G and 5G networks or other suitable radio access technologies, including IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20. The wireless communication system 100 may support radio access technologies beyond 5G, such as 6G. In addition, the wireless communication system 100 can support technologies such as Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), or Code Division Multiple Access (CDMA).

[0044] NE 102 may include application-enabled, vertical-enabled, or edge-enabled entities (SEAL, EDGEAPP, CAPIF) as specified in 3GPP SA6. One or more NE 102s may be distributed throughout a geographic area to form a wireless communication system 100. One or more of the NE 102s described herein may be, include, or be referred to as a network node, base station, network element, network function, network entity, wireless access network (RAN), NodeB, eNodeB (eNB), next-generation NodeB (gNB), or other suitable terms. NE 102 and UE 104 may communicate via a communication link, which may be wireless or wired. For example, NE 102 and UE 104 may perform wireless communication (e.g., receive signaling, transmit signaling) through a Uu interface.

[0045] NE 102 can provide a geographic coverage area, whereby NE 102 can support services for one or more UE 104s within the geographic coverage area. For example, NE 102 and UE 104 can support wireless communication of signals associated with services (e.g., voice, video, packet data, messaging, broadcasting, etc.) using one or more radio access technologies. In some embodiments, NE 102 can be mobile, such as a satellite associated with a non-terrestrial network (NTN). In some embodiments, different geographic coverage areas associated with the same or different radio access technologies can overlap, but different geographic coverage areas can be associated with different NE 102s.

[0046] One or more UEs 104 may be distributed throughout the geographic area of ​​the wireless communication system 100. UE 104 may include or be referred to as a remote unit, mobile device, wireless device, remote device, subscriber device, transmitter device, receiver device, or some other suitable term. In some embodiments, UE 104 may be referred to as a unit, station, terminal, or client, and other instances thereof. Alternatively or additionally, UE 104 may be referred to as an Internet of Things (IoT) device, Internet of Everything (IoE) device, or Machine-Type Communication (MTC) device, and other instances thereof.

[0047] UE 104 can support direct wireless communication with other UE 104 via a communication link. For example, UE 104 can support direct wireless communication with another UE 104 via a device-to-device (D2D) communication link. In some implementations, such as vehicle-to-vehicle (V2V) deployments, vehicle-to-everything (V2X) deployments, or cellular-V2X deployments, the communication link may be referred to as a sidelink. For example, UE 104 can support direct wireless communication with another UE 104 via a PC5 interface.

[0048] NE 102 may support communication with CN 106 or another NE 102, or both. For example, NE 102 may interface with other NE 102 or CN 106 via one or more backhaul links (e.g., S1, N2, N2, or network interfaces). In some implementations, NE 102 may communicate directly with each other. In some other implementations, NE 102 may communicate with each other or indirectly (e.g., via CN 106). In some implementations, one or more NE 102 may include sub-components, such as access network entities, which may be instances of Access Node Controllers (ANCs). The ANC may communicate with one or more UE 104s via one or more other access network transmitting entities, which may be referred to as radio headends, smart radio headends, or transmit-receive points (TRPs).

[0049] CN 106 can support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. CN 106 can be an evolved packet core (EPC) or a 5G core (5GC), which may include control plane entities that manage access and mobility (e.g., Mobility Management Entity (MME), Access and Mobility Management Function (AMF)) and user plane entities that route packets to or interconnect to external networks (e.g., Serving Gateway (S-GW), Packet Data Network (PDN) Gateway (P-GW), or User Plane Function (UPF)). In some implementations, the control plane entities may manage non-access stratum (NAS) functions, such as mobility, authentication, and bearer management (e.g., data bearers, signaling bearers, etc.) of one or more UEs 104 served by one or more NEs 102 associated with CN 106.

[0050] CN 106 can communicate with the packet data network (e.g., via S1, N2, N2, or another network interface) through one or more backhaul links. The packet data network may contain an application server. In some implementations, one or more UEs 104 can communicate with the application server. UE 104 can establish a session (e.g., a Protocol Data Unit (PDU) session) with CN 106 via NE 102. CN 106 can use the established session (e.g., an established PDU session) to route services (e.g., control information, data, etc.) between UE 104 and the application server. A PDU session can be an instance of a logical connection between UE 104 and CN 106 (e.g., one or more network functions of CN 106).

[0051] In the wireless communication system 100, NE 102 and UE 104 can use the resources of the wireless communication system 100 (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers)) to perform various operations (e.g., wireless communication). In some embodiments, NE 102 and UE 104 may support different resource structures. For example, NE 102 and UE 104 may support different frame structures. In some embodiments, such as in 4G, NE 102 and UE 104 may support a single frame structure. In some other embodiments, such as in 5G and other suitable radio access technologies, NE 102 and UE 104 may support various frame structures (i.e., multiple frame structures). NE 102 and UE 104 may support various frame structures based on one or more sets of parameters.

[0052] The wireless communication system 100 may support one or more parameter sets, and the parameter sets may include subcarrier spacing and cyclic prefixes. A first parameter set (e.g., μ=0) may be associated with a first subcarrier spacing (e.g., 15 kHz) and a regular cyclic prefix. In some embodiments, the first parameter set (e.g., μ=0) associated with the first subcarrier spacing (e.g., 15 kHz) may utilize one time slot per subframe. A second parameter set (e.g., μ=1) may be associated with a second subcarrier spacing (e.g., 30 kHz) and a regular cyclic prefix. A third parameter set (e.g., μ=2) may be associated with a third subcarrier spacing (e.g., 60 kHz) and a regular cyclic prefix or an extended cyclic prefix. A fourth parameter set (e.g., μ=3) may be associated with a fourth subcarrier spacing (e.g., 120 kHz) and a regular cyclic prefix. A fifth parameter set (e.g., μ=4) may be associated with a fifth subcarrier spacing (e.g., 240 kHz) and a regular cyclic prefix.

[0053] Time intervals for resources (such as communication resources) can be organized according to frames (also called radio frames). Each frame may have a duration, such as 10 milliseconds (ms). In some embodiments, each frame may contain multiple subframes. For example, each frame may contain 10 subframes, and each subframe may have a duration, such as 1 ms. In some embodiments, each frame may have the same duration. In some embodiments, each subframe of a frame may have the same duration.

[0054] Alternatively, the time intervals of resources (e.g., communication resources) can be organized according to time slots. For example, a subframe may contain a certain number (e.g., a certain quantity) of time slots. The number of time slots in each subframe may also depend on one or more parameter sets supported in the wireless communication system 100. For example, the first, second, third, fourth, and fifth parameter sets (i.e., μ=0, μ=1, μ=2, μ=3, μ=4) associated with corresponding subcarrier intervals of 15 kHz, 30 kHz, 60 kHz, 120 kHz, and 240 kHz can respectively utilize one time slot per subframe, two time slots per subframe, four time slots per subframe, eight time slots per subframe, and 16 time slots per subframe. Each time slot may contain a certain number (e.g., a certain quantity) of symbols (e.g., OFDM symbols). In some embodiments, the number (e.g., quantity) of time slots in a subframe may depend on the parameter set. For a conventional cyclic prefix, a time slot may contain 14 symbols. For an extended cyclic prefix (e.g., applicable to a 60 kHz subcarrier spacing), a time slot may contain 12 symbols. The relationship between the number of symbols per time slot, the number of time slots per subframe, and the number of time slots per frame for the regular and extended cyclic prefixes may depend on the parameter set. It should be understood that a reference to the first parameter set (e.g., μ=0) associated with the first subcarrier spacing (e.g., 15 kHz) may be used interchangeably between subframes and time slots.

[0055] In the wireless communication system 100, the electromagnetic (EM) spectrum can be divided into various categories, frequency bands, channels, etc., based on frequency or wavelength. For example, the wireless communication system 100 may support one or more operating frequency bands, such as frequency ranges represented as FR1 (410 MHz to 7.125 GHz), FR2 (24.25 GHz to 52.6 GHz), FR3 (7.125 GHz to 24.25 GHz), FR4 (52.6 GHz to 114.25 GHz), FR4a or FR4-1 (52.6 GHz to 71 GHz), and FR5 (114.25 GHz to 300 GHz). In some embodiments, NE 102 and UE 104 may perform wireless communication on one or more of the operating frequency bands. In some embodiments, FR1 may be used by NE 102 and UE 104, as well as other equipment or devices, for cellular communication services (e.g., control information, data). In some implementations, FR2 can be used by NE 102 and UE 104, as well as other equipment or devices, for short-range, high data rate capabilities.

[0056] FR1 may be associated with one or more parameter sets (e.g., at least three parameter sets). For example, FR1 may be associated with: a first parameter set (e.g., μ=0) containing a 15 kHz subcarrier spacing; a second parameter set (e.g., μ=1) containing a 30 kHz subcarrier spacing; and a third parameter set (e.g., μ=2) containing a 60 kHz subcarrier spacing. FR2 may be associated with one or more parameter sets (e.g., at least two parameter sets). For example, FR2 may be associated with: a third parameter set (e.g., μ=2) containing a 60 kHz subcarrier spacing; and a fourth parameter set (e.g., μ=3) containing a 120 kHz subcarrier spacing.

[0057] Figure 2 This document describes an example of an advanced transfer learning operation 200 based on aspects of this disclosure.

[0058] Transfer learning begins with a model 220 that has been previously trained on a large dataset for a specific task 210. After frequent training on a wide dataset, this model 220 has identified general features and patterns associated with many relevant tasks. This model 220 is called a pre-trained model or a pre-trained ML model.

[0059] The pre-trained model 220 is also called the base model. It consists of layers that have learned hierarchical feature representations using the incoming data. This base model forms the starting point for retraining for the second ML task 250.

[0060] In the pre-trained model 220, a set of layers is found that capture general information and knowledge 230 related to the new task 250 and the previous task 210. Because they tend to learn low-level information, these layers are often found near the top of the trained ML network.

[0061] Retraining selected layers using the dataset from the new task / challenge 250 is a procedure called fine-tuning. The goal is to preserve the knowledge from pre-training while enabling the new model 240 to modify its parameters to better suit the needs of the current assignment.

[0062] Figure 3 This describes an example of an advanced architecture 300 based on aspects of this disclosure. In the description... Figure 3 Before the advanced architecture 300, several roles will be introduced.

[0063] A pre-trained model producer (PMP) is an entity that provides ML models or ML model training entities that have already produced pre-trained ML models. This entity can be a Vertical Application Layer (VAL) server, an AI / ML Enabler server, or an Application Data Analytics Enabler (ADAE) server with Model Training and Management Entity (MTME) capabilities. In some embodiments, this entity is a service module, capability, or functionality of the aforementioned entities.

[0064] The Pre-trained Model Consumer (PMC) is an entity that consumes pre-trained ML models. This entity will use the pre-trained model to perform its own training on top and can enable VAL servers or further AIML servers or further ADAES or even 3GPP network analysis functions with MTME capabilities.

[0065] The Pretrained Model Registry (PMR) is optional and can be implemented as A-ADRF, an edge or cloud repository, or the CAPIF Core Functionality (CCF).

[0066] The TL Support Service Producer (TLSSP) can be the same entity as the PMC; however, it can differ, for example, when the PMC is a VAL Server (TLSSC) and it requires an AIML Enabler Server (TLSSP) to support the identification of pre-trained ML models for a given task. Therefore, the TLSSP tends to be applied when the pre-trained ML model training entity consuming the ML model is not the decision-making entity but delegates this to the enabler layer. This entity can be, for example, an AIML Enabler Server or an ADAES with MTME functionality.

[0067] A TL Support Service Consumer (TLSSC) is an entity whose role involves requesting support services to determine candidate and selected pre-trained ML models using specific criteria. This entity can be a VAL server or a 3GPP SA6 enabler functionality. This service is needed because this entity may not necessarily have the necessary permissions, capabilities, and knowledge to determine the best pre-trained ML model for a specific task, and it consumes services from an authorized producer.

[0068] Billing functionality (ChaF) is used for TLSS and / or for pre-trained ML model utilization (allowing CRUD operations in pre-trained ML models). This functionality can be part of the Operations, Administration, and Maintenance (OAM) layer or the Service Enabler Architecture Layer (SEAL) / or a part of the CAPIF platform provider.

[0069] about Figure 3 Example 300 illustrates a consumer of pre-trained model 310. The consumer of pre-trained model 310 can be a PMC or a TLSSC. Example 300 also illustrates a transfer learning enabler 320. The transfer learning enabler 320 can be a TLSSP. The transfer learning enabler 320 interfaces with model repository 330 (which can be a PMR), base model source 340 (which can be a PMP), enabler clients 382, ​​392 or UEs 380, 390, and ML enhancement application tasks 384 or 394. Example 300 also illustrates a ChaF 350 that interfaces with model repository 330, 5GC 360, and RAN 370. UEs 380 and 390 interface with RAN 370 via Uu. The various high-level procedural steps for transfer learning in Example 300 will now be described.

[0070] In the first step 301, TLSSC 310 detects the need to use transfer learning on a specific ML task / job ID or on a given analytics ID.

[0071] In a further step 302, TLSSC 310 subscribes to or requests TLSSP 320 to determine one or more available pre-trained ML models for a given ML task / job or analytics ID. In this step, TLSSC 310 provides all necessary contextual information, including feature sets, dataset requirements, model information, the required confidence level for the analyzed event, UEs or a list of UEs, services or service profiles, regions of interest, and timeframes.

[0072] In a further step 303, TLSSP 320 (acting as PMC 384) finds a list of PMR 330 or PMP 340 entities that provide pre-trained ML models in a given service area.

[0073] In the absence of PMR 330, entities providing pre-trained ML models need to be discovered at TLSSP 320 based on a list of pre-configured entities / platforms, or by querying PMP 340 that has previously provided pre-trained ML models for other tasks, or by querying PMP entities 340 with MTME functionality in other enablers.

[0074] In a further step 304, TLSSP 320 (acting as PMC 384) sends a request to PMR 330 to provide all pre-trained ML model information that is allowed to be opened. This request may be for all or a subset of ML models in a given region, or for a given analysis task, or for a given set of ML model task / job IDs, or for a given set of features / context information, or for a given service / service profile or a set of UEs.

[0075] In a further step 305, PMR 330 authorizes and can also interact with ChaF 350 to check the billing information of each in the pre-trained ML model.

[0076] In a further step 306, PMR 330 responds to TLSSP 320 (acting as PMC 384) with information about a list of pre-trained ML models based on the request.

[0077] In a further step 307, TLSSP 320 identifies similar models that can be used as pre-trained ML models for transfer learning and can also rate them and assess their applicability and potential benefits.

[0078] Criteria used for identification may include the following: similar features / data requirements to the requested task; similarity of tasks (e.g., similar analysis types); similarity of the ML methods used; whether the pre-trained ML model has been used in the past for tasks / purposes similar to the requested task; previous ratings / evaluations of the ML model based on feedback from other PMCs; and the reputation / rating of the source of the pre-trained model (i.e., the vendor, server, operator, edge provider).

[0079] In a further step 308, TLSSP 320 sends one or more proposed pre-trained ML model information to TLSSC 310, optionally with some ratings or some contextual information, to allow the consumer to select based on the proposals or simply notify the selected model.

[0080] Figure 4 This describes instance 400 of a TL-enabled message stream triggered by ADAE according to aspects of this disclosure.

[0081] In this example, the entities involved are those defined in 3GPP SA6TR 23.700-82 and SEAL ADES (specifically in 3GPP TS 23.434 and 3GPP TS 23.436, titled "Service enabler architecture layer for verticals (SEAL); functional architecture and information flows"). Before discussing message flows, a brief mapping to the aforementioned roles will be provided.

[0082] TLSSC 410 is an ADAES with MTME capability, which requires identifying where and how to obtain pre-trained ML models.

[0083] TLSSP 420 is a functional / service module used to provide TL-enabled AI / ML enabled servers or servers.

[0084] The PMP 440 can enable a VAL server or another AI / ML server or another platform capability with MTME capabilities (such as another ADAES).

[0085] PMC is the same as TLSSC or TLSSP and can enable servers or VAL servers for ADAES or AI / ML.

[0086] PMR 430 is an A-ADRF, CCF, or edge repository.

[0087] In the first step 401, the ADAES 410 with MTME capability sends a request to the AIML enabling server (with TL enabler capability) 420 to support the discovery and selection of an appropriate pre-trained ML model for the analytics ID. This request may include the analytics ID and the application service ID / profile applied thereto, a list of UEs / UEs applied thereto, the region and time of interest and validity, information to aid selection such as required features / data requirements, environment, confidence level, type of TL, method / algorithm to be used, licensing and limitations of the model to be selected (e.g., from other vendors, operators), KPIs for delivering the pre-trained model (e.g., maximum latency), and whether there is a preference for a list of ML models or sources / platforms to be used.

[0088] In a further step 402, an AI / ML enablement server 420 with TL enabler capabilities discovers a repository 430 or directly discovers a potential MTME source 440 that can provide a pre-trained ML model for this request. Such entities can be VAL servers from the same or other platforms, or other ADAEs or other AI / ML enablement servers.

[0089] Entities 430 and 440 can be discovered using CAPIF (which interacts with CCF to discover A-ADRF or other servers).

[0090] In a further step 403, the AI / ML enabling server 420 sends a request message to receive information about pre-trained ML models available in a given service area. This request message is sent to the repository 430 or directly to the source of the base model 440. This request may contain the criteria of step 401 or a subset thereof.

[0091] In a further step 404, the AI / ML enablement server 420 receives information about available pre-trained ML models and contextual information to assist in selection based on a request.

[0092] In a further step 405, the AI / ML enabling server 420 uses specific guidelines (as described above) with optional support for A-ADRF 430. Figure 3 (Description) to evaluate whether a pre-trained ML model is suitable for analytics ID. This can be aided by data / statistics from previous use of these models for a specific analytics type.

[0093] The AI / ML Enabled Server 420 can rate or weight pre-trained ML models or their sources. This rating can be used to assess the suitability of a specific analytics ID (i.e., <analysis ID, rating 90%>). For completeness, the rating can be in the form of a weight, percentage, relative / actual value, or qualitative parameter (i.e., "good", "average"), which is used to evaluate the model's suitability for transfer learning. This rating can be derived from feedback received from previous training from consumers or service providers or from further validation entities that validate the model or the rating of the source.

[0094] Based on the evaluation (which may be based on rating), the AI / ML Enabled Server 420 determines one or more pre-trained ML models for analysis ID.

[0095] In a further step 406, the AI / ML enabling server 420 sends information (ID, configuration file, download address) for the selected / candidate pre-trained ML model, and optionally information about the pre-trained ML model itself (if this entity has one), to ADAES 410. Furthermore, if ADAES 410 needs to make a decision within its list, this may include the rating / weight of the pre-trained ML model.

[0096] In a further step 407, ADAES 410 may extract the selected model from the repository 430 or directly from the corresponding MTME entity 440.

[0097] Figure 5 This describes instance 500 of a message stream enabled by a TL triggered by a VAL according to aspects of this disclosure.

[0098] In this example 500, the entities involved are those defined in 3GPP SA6 TR 23.700-82 and SEAL ADES (3GPP TS 23.434, 3GPP TS 23.436). Before discussing message flow, a brief overview of the mapping to the aforementioned roles will be provided.

[0099] TLSSC 510 is a VAL server that needs to identify where and how the pre-trained ML model is obtained.

[0100] TLSSP 520 is a functional / service module used to provide TL-enabled AI / ML enabled servers or servers.

[0101] The PMP 540 can enable a VAL server or another AI / ML server or another platform capability with MTME capabilities (such as another ADAES).

[0102] PMR 530 is an A-ADRF, CCF, or edge repository.

[0103] In the first step 501, the VAL server 510 sends a request to the AI / ML enabling server (with TL enabler capability) 520 to support the discovery and selection of an appropriate pre-trained ML model for a given ML model task or a given ML model profile. This request may include the task ID / profile or model profile and the application service ID / type applied thereto, a list of UEs / UEs applied thereto, the area of ​​interest and time and validity, information to aid selection such as required features / data requirements, environment, confidence level, type of TL, method / algorithm to be used, licensing and limitations of the model to be selected (e.g., from other vendors, operators), KPIs for delivering the pre-trained model (e.g., maximum latency), and whether there is a preference for a list of ML models or sources / platforms to be used. The ML model profile may be a set of attributes that identify similar models.

[0104] In a further step 502, an AI / ML enabler server 520 with TL enabler capabilities discovers a repository 530 or directly discovers a potential MTME source 540 that can provide a pre-trained ML model for this request. Such entities can be VAL servers from the same or other platforms, or other ADAE or other AIML enablers.

[0105] Step 502 also includes mapping the requested ML model profile to a list of applicable ML task IDs so that the availability of pre-trained ML models is requested only for the selected task list under this profile.

[0106] If no ML model profile is used, then AI / ML Enable Server 520 searches all ML models in a given service area or based on the requested features.

[0107] In a further step 503, the AI / ML enabling server 520 sends a request message to receive information about pre-trained ML models available in a given service region or under the same ML model profile or ML task list. This request message is sent to the repository 530 or directly to the source 540 of the underlying ML model. This request contains the criteria of step 501 or a subset thereof.

[0108] In a further step 504, the AI / ML enabling server 520 receives information about available pre-trained ML models and contextual information to assist in selection based on a request.

[0109] In a further step 505, the AI / ML Enabled Server 520, with optional support of A-ADRF 530, uses specific criteria (mentioned in the advanced solutions) to evaluate whether a pre-trained ML model is suitable for an ML model task or ML model profile. This can be aided by data / statistics from previous use of these models for a specific model profile or task.

[0110] AI / ML Enabled Server 520 can rate or set weights for pre-trained ML models or the sources of ML models. This rating can be used to assess the suitability of a specific analytics ID (i.e., <Model Profile ID, rating 90%> or <Task ID, rating 0.98>).

[0111] Based on the evaluation (which may be based on rating), the AI / ML Enabled Server 520 determines one or more candidate / selected pre-trained ML models for a task or profile.

[0112] In a further step 506, the AI / ML enabling server 520 sends information (ID, configuration file, download address) for selecting the pre-trained ML model, and optionally information about the pre-trained ML model itself (if this entity exists), to the VAL server 510. Furthermore, if the VAL server 510 needs to make a decision in its list, this may include the rating / weight of the pre-trained ML model.

[0113] In a further step 507, the VAL server 510 may retrieve the selected pre-trained ML model from the repository 530 or directly from the corresponding MTME entity 540.

[0114] Figure 6 An example of a UE 600 according to aspects of this disclosure is described. UE 600 may include a processor 602, a memory 604, a controller 606, and a transceiver 608. The processor 602, memory 604, controller 606, or transceiver 608, or various combinations thereof, or various components thereof, may be examples of components for performing the aspects of this disclosure as described herein. These components may be coupled via one or more interfaces (e.g., operative ground, communicative ground, functional ground, electronic ground, related electrical ground).

[0115] Processor 602, memory 604, controller 606, or transceiver 608, or various combinations or components thereof, may be implemented in hardware (e.g., a circuit system). The hardware may include processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), or other programmable logic devices, or any combination thereof, configured to or otherwise support components for performing the functions described in this disclosure.

[0116] Processor 602 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, ASICs, FPGAs, or any combination thereof). In some embodiments, processor 602 may be configured to operate memory 604. In some other embodiments, memory 604 may be integrated into processor 602. Processor 602 may be configured to execute computer-readable instructions stored in memory 604 to cause UE 600 to perform various functions of this disclosure.

[0117] Memory 604 may include volatile or non-volatile memory. Memory 604 may store computer-readable, computer-executable code containing instructions that, when executed by processor 602, cause UE 600 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium, such as memory 604 or another type of memory. Computer-readable medium includes both non-transitory computer storage media and communication media, including any media that facilitates the transfer of computer programs from one place to another. Non-transitory storage media may be any available media accessible by a general-purpose or special-purpose computer.

[0118] In some embodiments, processor 602 and memory 604 coupled to processor 602 may be configured to cause UE 600 to perform one or more of the functions described herein (e.g., instructions stored in memory 604 are executed by processor 602). For example, processor 602 may support wireless communication at UE 600 according to examples disclosed herein. UE 600 may be configured to support components for performing aspects of the methods disclosed herein. UE 600 may be... Figure 3 UE 380 or 390.

[0119] Controller 606 manages the input and output signals of UE 600. Controller 606 can also manage peripheral devices not integrated into UE 600. In some embodiments, controller 606 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some embodiments, controller 606 may be implemented as part of processor 602.

[0120] In some embodiments, UE 600 may include at least one transceiver 608. In other embodiments, UE 600 may have more than one transceiver 608. Transceiver 608 may represent a wireless transceiver. Transceiver 608 may include one or more receiver chains 610, one or more transmitter chains 612, or a combination thereof.

[0121] Receiver chain 610 may be configured to receive signals (e.g., control information, data, packets) via wireless media. For example, receiver chain 610 may include one or more antennas for receiving signals via air or wireless media. Receiver chain 610 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. Receiver chain 610 may include at least one demodulator configured to demodulate the received signal and obtain transmitted data by reversing the modulation technique applied during signal transmission. Receiver chain 610 may include at least one decoder for decoding the demodulated signal to receive transmitted data.

[0122] Transmitter chain 612 can be configured to generate and transmit signals (e.g., control information, data, packets). Transmitter chain 612 may include at least one modulator for modulating data onto a carrier signal to prepare the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques, such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase shift keying (PSK) or quadrature amplitude modulation (QAM). Transmitter chain 612 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over a wireless medium. Transmitter chain 612 may also include one or more antennas for transmitting the amplified signal into the air or a wireless medium.

[0123] Figure 7 An example of a processor 700 according to aspects of this disclosure is described. Processor 700 may be an example of a processor configured to perform various operations according to the examples described herein. Processor 700 may include a controller 702 configured to perform various operations according to the examples described herein. Processor 700 may optionally include at least one memory 704, which may be, for example, an L1 / L2 / L3 cache. Additionally or alternatively, processor 700 may optionally include one or more arithmetic logic units (ALUs) 706. One or more of these components may be electronically communicated or otherwise coupled (e.g., operatively, communicatively, functionally, electronically, or relatingly) via one or more interfaces (e.g., buses).

[0124] Processor 700 may be a processor chipset and includes a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receive, acquire, retrieve, transmit, output, forward, store, determine, identify, access, write, read) according to the examples described herein. The processor chipset may include one or more cores, one or more caches (e.g., memory local to the processor chipset (e.g., processor 700) or contained within the processor chipset (e.g., processor 700)) or other memory (e.g., random access memory (RAM), read-only memory (ROM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), ferroelectric RAM (FeRAM), magnetic RAM (MRAM), resistive RAM (RRAM), flash memory, phase-change memory (PCM), etc.).

[0125] Controller 702 can be configured to manage and coordinate various operations of processor 700 (e.g., signaling, receiving, acquiring, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, and reading) to enable processor 700 to support various operations according to the examples described herein. For example, controller 702 can operate as a control unit of processor 700 to generate control signals that manage the operation of various components of processor 700. These control signals include enabling or disabling functional units, selecting data paths, initiating memory access, and coordinating the timing of operations.

[0126] Controller 702 may be configured to fetch (e.g., fetch, retrieve, receive) instructions from memory 704 and determine subsequent instructions to be executed to enable processor 700 to support various operations according to examples described herein. Controller 702 may be configured to track the memory addresses of instructions associated with memory 704. Controller 702 may be configured to decode instructions to determine the operations to be performed and the operands involved. For example, controller 702 may be configured to interpret instructions and determine control signals output to other components of processor 700 to enable processor 700 to support various operations according to examples described herein. Alternatively or additionally, controller 702 may be configured to manage data flow within processor 700. Controller 702 may be configured to control data transfers between registers, arithmetic logic unit (ALU), and other functional units of processor 700.

[0127] Memory 704 may include one or more caches (e.g., memory local to or included in processor 700) or other memories, such as RAM, ROM, DRAM, SDRAM, SRAM, MRAM, flash memory, etc. In some embodiments, memory 704 may reside within or on the processor chipset (e.g., locally to processor 700). In some other embodiments, memory 704 may reside outside the processor chipset (e.g., remotely from processor 700).

[0128] Memory 704 may store computer-readable, computer-executable code containing instructions that, when executed by processor 700, cause processor 700 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium, such as system memory or another type of memory. Controller 702 and / or processor 700 may be configured to execute computer-readable instructions stored in memory 704 to cause processor 700 to perform various functions. For example, processor 700 and / or controller 702 may be coupled to or coupled to memory 704, and processor 700, controller 702, and memory 704 may be configured to perform the various functions described herein. In some instances, processor 700 may include multiple processors and memory 704 may include multiple memories. One or more of the multiple processors may be coupled to one or more of the multiple memories, which may be individually or jointly configured to perform the various functions described herein.

[0129] One or more ALU 706s may be configured to support various operations according to the examples described herein. In some embodiments, one or more ALU 706s may reside within or on a processor chipset (e.g., processor 700). In some other embodiments, one or more ALU 706s may reside outside the processor chipset (e.g., processor 700). One or more ALU 706s may perform one or more calculations on data, such as addition, subtraction, multiplication, and division. For example, one or more ALU 706s may receive input operands and operation codes that determine the operation to be performed. One or more ALU 706s may be configured with various logic and arithmetic circuitry (including adders, subtractors, shifters, and logic gates) to process and manipulate data according to the operation. Alternatively, one or more ALU 706s may support logical operations such as AND, OR, XOR, NOR, and NAND to enable one or more ALU 706s to handle conditional operations, comparisons, and bitwise operations.

[0130] Processor 700 may support wireless communication according to examples disclosed herein. Processor 700 may be configured or operable to support components for performing aspects of the methods disclosed herein.

[0131] Figure 8 This describes an example of an NE 800 according to aspects of this disclosure. The NE 800 may include an application-enabled, vertical-enabled, or edge-enabled entity (SEAL, EDGEAPP, CAPIF) specified in 3GPP SA6. The NE 800 may include a processor 802, a memory 804, a controller 806, and a transceiver 808. The processor 802, memory 804, controller 806, or transceiver 808, or various combinations thereof, or various components thereof, may be examples of components for performing the aspects of this disclosure as described herein. These components may be coupled via one or more interfaces (e.g., operative ground, communication ground, functional ground, electronic ground, associated electrical ground).

[0132] Processor 802, memory 804, controller 806, or transceiver 808, or various combinations or components thereof, may be implemented in hardware (e.g., a circuit system). The hardware may include processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), or other programmable logic devices, or any combination thereof, configured to or otherwise support components for performing the functions described in this disclosure.

[0133] Processor 802 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, ASICs, FPGAs, or any combination thereof). In some embodiments, processor 802 may be configured to operate memory 804. In some other embodiments, memory 804 may be integrated into processor 802. Processor 802 may be configured to execute computer-readable instructions stored in memory 804 to cause NE 800 to perform various functions of this disclosure.

[0134] Memory 804 may include volatile or non-volatile memory. Memory 804 may store computer-readable, computer-executable code containing instructions that, when executed by processor 802, cause NE 800 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium, such as memory 804 or another type of memory. Computer-readable medium includes both non-transitory computer storage media and communication media, including any media that facilitates the transfer of computer programs from one place to another. Non-transitory storage media may be any available media accessible by a general-purpose or special-purpose computer.

[0135] In some embodiments, processor 802 and memory 804 coupled to processor 802 may be configured such that NE 800 performs one or more of the functions described herein (e.g., instructions stored in memory 804 are executed by processor 802). For example, processor 802 may support wireless communication at NE 800 according to examples disclosed herein. NE 800 may be configured to support components for performing aspects of the methods disclosed herein. NE 800 may be... Figure 3 The NE series includes 310, 320, 330, 340, 350, 360, and 370. The NE 800 is also available. Figure 4 The NE 410, 420, 430, and 440 are mentioned. The NE 800 is also mentioned. Figure 5 NE 510, 520, 530, 540.

[0136] Controller 806 manages the input and output signals of NE 800. Controller 806 can also manage peripheral devices not integrated into NE 800. In some embodiments, controller 806 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some embodiments, controller 806 may be implemented as part of processor 802.

[0137] In some embodiments, NE 800 may include at least one transceiver 808. In other embodiments, NE 800 may have more than one transceiver 808. Transceiver 808 may represent a wireless transceiver. Transceiver 808 may include one or more receiver chains 810, one or more transmitter chains 812, or a combination thereof.

[0138] Receiver chain 810 may be configured to receive signals (e.g., control information, data, packets) via wireless media. For example, receiver chain 810 may include one or more antennas for receiving signals via air or wireless media. Receiver chain 810 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. Receiver chain 810 may include at least one demodulator configured to demodulate the received signal and obtain transmitted data by reversing the modulation technique applied during signal transmission. Receiver chain 810 may include at least one decoder for decoding the demodulated signal to receive transmitted data.

[0139] Transmitter chain 812 can be configured to generate and transmit signals (e.g., control information, data, packets). Transmitter chain 812 may include at least one modulator for modulating data onto a carrier signal to prepare the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques, such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase shift keying (PSK) or quadrature amplitude modulation (QAM). Transmitter chain 812 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over a wireless medium. Transmitter chain 812 may also include one or more antennas for transmitting the amplified signal into the air or a wireless medium.

[0140] Figure 9A flowchart illustrating a method according to an aspect of this disclosure is provided. The operation of the method may be implemented by an NE as described herein. In some embodiments, the NE may execute a set of instructions to control the functional elements of the NE to perform the described functions.

[0141] At 902, the method may include receiving from a second network entity a first request to identify at least one pre-trained machine learning 'ML' model for transfer learning tasks, wherein the first request includes one or more criteria for identifying at least one pre-trained ML model. Operation 902 may be performed according to examples as described herein. In some embodiments, aspects of operation 902 may be provided by reference to [reference needed]. Figure 8 The described NE execution.

[0142] In operation 904, the method may include discovering a third network entity based on one or more criteria to indicate the availability of a pre-trained ML model. Operation 904 may be performed according to the examples described herein. In some implementations, aspects of operation 904 may be derived from references... Figure 8 The described NE execution.

[0143] In operation 906, the method may include obtaining first information about one or more available pre-trained ML models from a third network entity based on one or more criteria. Operation 906 may be performed according to the examples described herein. In some implementations, aspects of operation 906 may be derived from references... Figure 8 The described NE execution.

[0144] At 908, the method may include providing second information to a second network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for transfer learning for ML model tasks. Operation 908 may be performed according to the examples described herein. In some embodiments, aspects of operation 908 may be provided by reference to [reference needed]. Figure 8 The described NE execution.

[0145] Figure 10 A flowchart illustrating a method according to an aspect of this disclosure is provided. The operation of the method may be implemented by an NE as described herein. In some embodiments, the NE may execute a set of instructions to control the functional elements of the NE to perform the described functions.

[0146] In 1002, the method may include sending a first request to a first network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying at least one pre-trained ML model. Operation 1002 may be performed according to instances as described herein. In some embodiments, aspects of operation 1002 may be as referenced... Figure 8 The described NE execution.

[0147] At 1004, the method may include receiving second information from a first network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for transfer learning of ML model tasks. Operation 1004 may be performed according to instances as described herein. In some implementations, aspects of operation 1004 may be provided by reference to [reference needed]. Figure 8 The described NE execution.

[0148] Figure 11 A flowchart illustrating a method according to an aspect of this disclosure is provided. The operation of the method may be implemented by an NE as described herein. In some embodiments, the NE may execute a set of instructions to control the functional elements of the NE to perform the described functions.

[0149] At 1102, the method may include receiving a second request from a first network entity for an available pre-trained ML model, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model. Operation 1102 may be performed according to examples as described herein. In some embodiments, aspects of operation 1102 may be provided by reference to [reference needed]. Figure 8 The described NE execution.

[0150] In step 1104, the method may include determining first information about one or more available pre-trained ML models based on one or more criteria. Operation 1104 may be performed according to instances as described herein. In some implementations, aspects of operation 1104 may be derived from references... Figure 8 The described NE execution.

[0151] At 1106, the method may include sending a second response, comprising first information including one or more available pre-trained ML models, to a first network entity. Operation 1106 may be performed according to instances as described herein. In some implementations, aspects of operation 1106 may be as referenced... Figure 8 The described NE execution.

[0152] It should be noted that the methods described herein describe possible implementations, and the operations and steps may be rearranged or otherwise modified, and other implementations are possible.

[0153] This disclosure provides a first network entity for wireless communication, comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the first network entity: receives from a second network entity a first request to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; discovers a third network entity based on the one or more criteria for indicating the availability of the pre-trained ML model; obtains first information from the third network entity regarding one or more available pre-trained ML models based on the one or more criteria; and provides second information to the second network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for transfer learning for an ML model task.

[0154] Transfer learning is a technique in machine learning where a model trained on a first ML task is used as a starting point for a model on a second ML task. This is useful when the second ML task is similar to the first and / or when there is limited data available to train a completely new ML model for the second task. By using features learned from the first ML task as a starting point, the ML model can learn faster and more efficiently on the second ML task. This also helps prevent overfitting, as the ML model has learned general features that may be useful in the second ML task.

[0155] A pre-trained ML model is an ML model that has been previously trained for a specific ML task. Pre-trained ML models may have been trained on a wide range of training datasets. They may have identified general features and patterns relevant to many related ML tasks. A pre-trained ML model can also be called a base ML model. It can consist of layers that have learned hierarchical feature representations using training data from a first ML task. In transfer learning, a set of layers from the base model that captures general information relevant to a second ML task can be identified. Using the available training data from the second ML task, selected layers of the base model can be retrained—a process called 'fine-tuning'. This preserves the knowledge gained from pre-training (for the first ML task) while allowing the ML model to be modified to better suit the needs of the second ML task.

[0156] 3GPP technical specification 22.261, titled "Service Requirements for the 5G System," states that 5G systems need to support the distribution and sharing of AI / ML models / data. A key issue then becomes how to support transfer learning at the application enablement layer. More specifically, this raises questions about how to identify whether a pre-trained model can be used for another ML task and how to ensure that this solution will be beneficial. Further questions arise about how to configure transfer learning enablement entities to distribute pre-trained models.

[0157] The current 3GPP solution does not consider transfer learning technology, and therefore does not support the identification and selection of transfer learning entities and pre-trained models.

[0158] This disclosure provides for the identification of available pre-trained models for transfer learning tasks in ML models. This disclosure further provides criteria for selecting available pre-trained models. This disclosure further provides the entities involved in configuring the distribution of pre-trained ML models for transfer learning (i.e., the source of the pre-trained models, the destination of the pre-trained models, and the coordinator / registry). Licensing and billing aspects are also provided.

[0159] The solution described in this paper utilizes a middleware transfer learning support service producer (TLSSP) to provide identification and selection of pre-trained ML models for ML modeling tasks.

[0160] The solutions described in this paper provide enhancements to existing application-layer registries for ML models. Such solutions enable the benefits of transfer learning within 3GPP systems.

[0161] The first, second, and third network entities may belong to the wireless communication system. For example, the first network entity may be an application entity of the wireless communication system, the second network entity may be an application entity of the wireless communication system, and / or the third network entity may be an application entity of the wireless communication system.

[0162] The third network entity can be a pre-trained model producer (PMP). This entity can be an ML model provider or an ML model training entity that has already produced a pre-trained model. Instances of this entity include VAL servers, AIML-enabled servers, or ADAES with MTME capabilities. This entity can be a service module, capability, or functionality of the aforementioned entities.

[0163] The third network entity can be a pre-trained model registry (PMR). This registry can be optional and can be implemented as A-ADRF, an edge or cloud repository, or the CAPIF core functionality (CCF).

[0164] This article also introduces the Pre-trained Model Consumer (PMC). This entity consumes a pre-trained ML model. This entity will use the pre-trained model to perform its own training on top and can operate as a VAL server, a further AIML enable server, or a further ADAES or even 3GPP network analysis function with MTME capabilities.

[0165] The first network entity can be a Transfer Learning Support Service Producer (TLSSP). A TLSSP can be the same entity as a PMC (however, it can differ, for example, if the PMC is a VAL Server (TLSSC) and it requires an AIML Enabled Server (TLSSP) to support the identification of pre-trained models for a given task). A TLSSP can be, for example, an AIML Enabled Server or an ADAES with MTME functionality.

[0166] The second network entity may be referred to herein as the Transfer Learning Support Service Consumer (TLSSC). The TLSSC has the function of requesting support services (TLSSP) to determine candidate available pre-trained ML models and selecting a pre-trained ML model using specific criteria. This entity can operate as a VAL server. This entity can be a 3GGP SA6 enabler function. The reason for needing this service is that this entity does not have the necessary permissions, capabilities, or knowledge to determine the best pre-trained model for a specific task and consume services from an authorized producer (TLSSP) for selection.

[0167] This article also mentions the Billing Functionality (ChaF) for TLSS and / or pre-trained model utilization (which allows CRUD operations in the pre-trained model). This functionality can be part of OAM or part of the SEAL / EDGE / CAPIF platform provider.

[0168] The second information may be at least partially based on the first information. The receipt of the first request may be referred to herein as the receipt of the application requirement.

[0169] One or more criteria may include contextual information for identifying at least one pre-trained ML model, wherein the contextual information includes at least one of the following: one or more required features of the pre-trained ML model; one or more training data requirements; confidence required for analysis; required type of transfer learning; one or more user equipment (UE) associated with the ML model task; service or service type associated with the ML model task; region of interest and / or time associated with the ML model task; environment associated with the ML model task; required type of ML algorithm; one or more licenses for the pre-trained ML model; one or more restrictions on the pre-trained ML model; and key performance indicators (KPIs) for delivering the pre-trained ML model.

[0170] One or more criteria may include at least one of the following: an identifier for the analysis; an identifier for the ML model task; an identifier for the application service; an identifier for the service profile; an ML model profile; or an ML model task profile.

[0171] One or more criteria may include an ML model profile, wherein at least one processor may be configured to enable a first network entity to discover a third network entity based on one or more criteria by causing the first network entity to: map the ML model profile to a list of identifiers for an ML model task; and discover the third network entity based on the mapping.

[0172] One or more criteria may include an ML model task profile, wherein at least one processor may be configured to enable a first network entity to discover a third network entity based on one or more criteria by causing the first network entity to: map the ML model task profile to a list of identifiers of a pre-trained ML model; and discover the third network entity based on the mapping.

[0173] At least one processor may be configured to cause the first network entity to obtain first information from a third network entity by causing the first network entity to: send a second request for an available pre-trained ML model to the third network entity, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; and receive a second response from the third network entity including the first information.

[0174] At least one processor may be configured to enable a first network entity to evaluate one or more available pre-trained ML models as pre-trained ML models for ML modeling tasks.

[0175] At least one processor may be configured to cause the first network entity to evaluate one or more available pre-trained ML models by rating one or more available pre-trained ML models, thereby generating rating parameters.

[0176] At least one processor may be configured to enable a first network entity to rate one or more available pre-trained ML models based on one or more similarities to the ML model task, wherein the one or more similarities include those based on: one or more features of the pre-trained ML model; training data requirements of the pre-trained ML model; ML model algorithm of the pre-trained ML model; previous use of the pre-trained ML model; previous ratings of the pre-trained ML model; and ratings of the source of the pre-trained ML model.

[0177] At least one processor may be configured to enable a first network entity to rate one or more available pre-trained ML models for the applicability of the following identifiers: analysis; ML model task; or ML model profile.

[0178] The second piece of information may include rating parameters.

[0179] The second piece of information may include one or more available pre-trained ML models.

[0180] The first network entity may include an AIML (Artificial Intelligence Machine Learning) enabling server. The second network entity may include an ADAES (Application Data Analytics Enabler Server), an edge application server, or a VAL (Vertical Application Layer) server. The third network entity may include a PMR (Pre-trained Model Repository) or a PMP (Pre-trained Model Provider), wherein optionally: PMR is an AADRF (Application Layer Analytics Data Repository Function), a CCF (Common API Framework Core Function), an ADRF (Analytical Data Repository Function), or an edge repository; and / or PMP is a VAL server, an AIML enabling server, or a MTME (Model Training and Management Entity).

[0181] ML modeling tasks can be analytical tasks.

[0182] A second network entity for wireless communication is provided, comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the second network entity: sends a first request to a first network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying at least one pre-trained ML model; and receives second information from the first network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for transfer learning used for an ML model task.

[0183] The first network entity may include an AIML (Artificial Intelligence Machine Learning) enabling server. The second network entity may include an ADAES (Application Data Analytics Enabler) server or a VAL (Vertical Application Layer) server.

[0184] One or more criteria may include contextual information for identifying at least one pre-trained ML model, wherein the contextual information includes at least one of the following: one or more required features of the pre-trained ML model; one or more training data requirements; confidence required for analysis; required type of transfer learning; one or more user equipment (UE) associated with the ML model task; service or service type associated with the ML model task; region of interest and / or time associated with the ML model task; environment associated with the ML model task; required type of ML algorithm; one or more licenses for the pre-trained ML model; one or more restrictions on the pre-trained ML model; and key performance indicators (KPIs) for delivering the pre-trained ML model.

[0185] One or more criteria may include at least one of the following: an identifier for the analysis; an identifier for the ML model task; an identifier for the application service; an identifier for the service profile; an ML model profile; or an ML model task profile.

[0186] The second piece of information may include rating parameters.

[0187] The second piece of information may include one or more available pre-trained ML models.

[0188] At least one processor can be configured to enable a second network entity to extract one or more available pre-trained ML models (e.g., from PMR or PMP).

[0189] At least one processor may be configured to enable a second network entity to determine the need to use transfer learning for a specific ML task, a specific job identifier, or a specific analytics identifier. In response to determining this need, the second network entity may send a first request.

[0190] A third network entity for wireless communication is provided, comprising: at least one memory; and at least one processor coupled to the at least one memory and configured such that the third network entity: receives from a first network entity a second request for an available pre-trained ML model, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; determines first information of one or more available pre-trained ML models based on the one or more criteria; and sends a second response including the first information of one or more available pre-trained ML models to the first network entity.

[0191] The first network entity may include an AIML (Artificial Intelligence Machine Learning) enabling server. The third network entity may include a pre-trained model repository 'PMR' or a pre-trained model provider 'PMP', wherein optionally: PMR is an application-layer analytics data repository function 'AADRF', a general API framework core function 'CCF', or an edge repository; and / or PMP is a VAL server, an AIML enabling server, or a model training and management entity 'MTME'.

[0192] One or more criteria may include contextual information for identifying at least one pre-trained ML model, wherein the contextual information includes at least one of the following: one or more required features of the pre-trained ML model; one or more training data requirements; confidence required for analysis; required type of transfer learning; one or more user equipment (UE) associated with the ML model task; service or service type associated with the ML model task; region of interest and / or time associated with the ML model task; environment associated with the ML model task; required type of ML algorithm; one or more licenses for the pre-trained ML model; one or more restrictions on the pre-trained ML model; and key performance indicators (KPIs) for delivering the pre-trained ML model.

[0193] One or more criteria may include at least one of the following: an identifier for the analysis; an identifier for the ML model task; an identifier for the application service; an identifier for the service profile; an ML model profile; or an ML model task profile.

[0194] A method is provided performed by a first network entity, comprising: receiving from a second network entity a first request to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being for an ML model task, wherein the first request includes one or more criteria for identifying at least one pre-trained ML model; discovering a third network entity based on the one or more criteria for indicating the availability of the pre-trained ML model; obtaining first information about one or more available pre-trained ML models from the third network entity based on the one or more criteria; and providing second information to the second network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for transfer learning for an ML model task.

[0195] The second information may be at least partially based on the first information.

[0196] The first network entity may include an AIML (Artificial Intelligence and Machine Learning) enabler server. The second network entity may include an ADAES (Application Data Analytics Enabler Server) or a VAL (Vertical Application Layer) server. The third network entity may include a PMR (Pre-trained Model Repository) or a PMP (Pre-trained Model Provider), wherein optionally: PMR is an AADRF (Application Layer Analytics Data Repository Function), a CCF (Common API Framework Core Function), or an edge repository; and PMP is a VAL server, an AIML enabler server, or a MTME (Model Training and Management Entity).

[0197] One or more criteria may include contextual information for identifying at least one pre-trained ML model. Contextual information may include at least one of the following: one or more required features of the pre-trained ML model; one or more training data requirements; confidence levels required for analysis; the required type of transfer learning; one or more user equipment (UEs) associated with the ML model task; the service or service type associated with the ML model task; the region of interest and / or time associated with the ML model task; the environment associated with the ML model task; the required type of ML algorithm; one or more licenses for the pre-trained ML model; one or more restrictions on the pre-trained ML model; and key performance indicators (KPIs) for delivering the pre-trained ML model.

[0198] One or more criteria may include at least one of the following: an identifier for the analysis; an identifier for the ML model task; an identifier for the application service; an identifier for the service profile; an ML model profile; or an ML model task profile.

[0199] Discovering third network entities based on one or more criteria may include: mapping ML model profiles to a list of identifiers for ML model tasks; and discovering third network entities based on the mapping.

[0200] Discovering third network entities based on one or more criteria may include: mapping ML model task profiles to a list of identifiers in a pre-trained ML model; and discovering third network entities based on the mapping.

[0201] Obtaining the first information from the third network entity may include: sending a second request for an available pre-trained ML model to the third network entity, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; and receiving a second response from the third network entity that includes the first information.

[0202] Methods may include evaluating one or more available pre-trained ML models as pre-trained ML models for ML modeling tasks.

[0203] Evaluating one or more available pre-trained ML models may include rating one or more available pre-trained ML models, thereby generating rating parameters.

[0204] The rating may include rating one or more available pre-trained ML models based on one or more similarities to the ML model task, wherein the one or more similarities include based on the following: one or more features of the pre-trained ML model; the training data requirements of the pre-trained ML model; the ML model algorithm of the pre-trained ML model; previous use of the pre-trained ML model; previous ratings of the pre-trained ML model; and the rating of the source of the pre-trained ML model.

[0205] Ratings can include assessing the applicability of one or more available pre-trained ML models to the following identifiers: analytics; ML model tasks; or ML model profiles.

[0206] The second piece of information may include rating parameters.

[0207] The second piece of information may include one or more available pre-trained ML models.

[0208] ML modeling tasks can be analytical tasks.

[0209] A method performed by a second network entity is further provided, comprising: sending a first request to a first network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying at least one pre-trained ML model; and receiving second information from the first network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for transfer learning used for an ML model task.

[0210] A further method is provided, performed by a third network entity, comprising: receiving a second request for an available pre-trained ML model from a first network entity, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; determining first information of one or more available pre-trained ML models based on the one or more criteria; and sending a second response including the first information of one or more available pre-trained ML models to the first network entity.

[0211] Transfer learning is used to support entities to simplify the training process (i.e., if regular ML training is not possible or if there is insufficient training data and time). For TL-enabled entities in 3GPP systems that are more application-layer focused, the following questions arise: how to identify whether a pre-trained model can be used for another task (i.e., a new ML job) and how to ensure that this solution will be beneficial; and how to configure TL-enabled entities to support the distribution of pre-trained models.

[0212] This disclosure describes the identification of candidate pre-trained ML models and their criterion-based selection. It also describes the configuration of the entities involved in the distribution (i.e., the source, destination, coordinator / registry of the pre-trained ML models). Licensing and billing aspects are also included. Such solutions utilize middleware (TLSSP) to provide identification and selection of pre-trained ML models for the requested ML modeling task.

[0213] The current solution does not consider transfer learning techniques (i.e., using a pre-trained ML model from one task to another). Therefore, entity identification and selection, as well as pre-trained ML models, are not currently supported in the 3GPP system.

[0214] A method is provided for supporting the discovery of transfer learning capabilities for one or more ML model tasks (at, for example, a TL enabler entity), the method comprising: receiving from a first application entity an application requirement to identify at least one pre-trained ML model for the ML model task, wherein the application requirement includes a set of criteria for identifying at least one pre-trained ML model; discovering a second application entity to provide information on the availability of the pre-trained ML model based on the set of criteria of the application requirement; obtaining information about one or more pre-trained ML models from the second application entity based on the set of criteria of the application requirement; selecting at least one pre-trained ML model from the obtained set of pre-trained ML models (based on, for example, an evaluation of the obtained pre-trained ML model); and sending information about the at least one selected pre-trained ML model to the application.

[0215] ML model tasks can be analytical tasks identified by analytical identifiers.

[0216] Application requirements may include those described in the examples in this document.

[0217] Discovery may include: sending a request to receive information to a second application entity; and receiving a response from the second application entity.

[0218] The second application entity can be an ML model repository.

[0219] The second application entity can be a pre-trained ML model producer entity (i.e., VAL entity, ADAES, or MTME).

[0220] The method may further include rating the obtained pre-trained model based on a set of criteria.

[0221] A set of criteria may include one or more criteria as described in the examples in this document.

[0222] Application requirements may include ML model configuration files or ML model task configuration files.

[0223] The method may further include mapping ML model profiles or ML model task profiles to a set of ML model task identifiers.

[0224] The method may further include mapping ML model profiles or ML model task profiles to one or more pre-trained ML model identifiers.

[0225] It should be noted that the methods described herein describe possible implementations, and the operations and steps may be rearranged or otherwise modified, and other implementations are possible.

[0226] The description herein is provided to enable those skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the scope of this disclosure. Therefore, this disclosure is not limited to the examples and designs described herein, but should be given the broadest scope consistent with the principles and novel features disclosed herein.

[0227] As used herein, "ML model" may include a mathematical algorithm that can be "trained" with data and human expert input as examples to replicate decisions made by experts when given the same information (see 3GPP TS 28.105). As used herein, "ML model training" may include the ability of an ML training function or service to acquire data, run it through an ML model, derive associated losses, and adjust the parameter settings of that ML model based on the calculated losses (see 3GPP TS 28.105). As used herein, "ML model inference" may include the ability of an ML model inference function to perform inference using an ML model and / or an AI decision entity (see 3GPP TS 28.105). As used herein, "AI / ML enablement" may include an application enablement framework consisting of one or more AI / ML enabler capabilities based on an SA6 provider implementation. This capability may be deployed as an enablement layer server (e.g., SEAL or ADAES) or a client (or deployed within it). As used herein, the term "repository" is used interchangeably with the term "model repository." The term "model" is used interchangeably with the term "ML model."

[0228] The following abbreviations are used in this article: TL: Transfer Learning; MDA: Management Domain Analytics; AIML: Artificial Intelligence / Machine Learning; ADAES: Application Data Analytics Enabler Server; SEAL: Service Enabler Architecture Layer; NW: Network; MTME: Model Training and Management Entity; CCF: Core Functionality of CAPIF (Common API Framework); CRUD: Create-Read-Update-Delete; A-ADRF: Application Layer-Analysis Data Repository Functionality; PMP: Pre-trained Model Producer; PMC: Pre-trained Model Consumer; TLSSP: TL Supported Service Producer; TLSSC: TL Supported Service Consumer; and ChaF: Billing Functionality.

Claims

1. A first network entity for wireless communication, comprising: At least one memory; and At least one processor, coupled to the at least one memory and configured to enable the first network entity to: A first request is received from a second network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; Based on one or more of the criteria, a third network entity is discovered to indicate the availability of the pre-trained ML model; Based on the one or more criteria, first information for one or more usable pre-trained ML models is obtained from the third network entity; and The second information is provided to the second network entity, the second information indicating that the one or more available pre-trained ML models are used as pre-trained ML models for the transfer learning of the ML model task.

2. The first network entity of claim 1, wherein the one or more criteria include contextual information for identifying the at least one pre-trained ML model, wherein the contextual information includes at least one of the following: One or more desired features of a pre-trained ML model; One or more training data requirements; The required confidence level for analysis; The required types of transfer learning; One or more user equipment (UEs) associated with the ML model task; The service or service type associated with the ML model task; The region of interest and / or time associated with the ML model task; The environment associated with the ML model task; The required types of ML algorithms; One or more licenses for a pre-trained ML model; One or more restrictions on pre-trained ML models; Key performance metrics (KPIs) used for delivering pre-trained ML models.

3. The first network entity according to any of the preceding claims, wherein the one or more criteria include at least one of the following: Identifier for analysis; The identifier of the ML model task; Identifier for application services; Identifier for the service configuration file; ML model configuration file; or ML model task configuration file.

4. The first network entity of claim 3, wherein the one or more criteria include the ML model profile, wherein the at least one processor is configured to cause the first network entity to discover the third network entity based on the one or more criteria by causing the first network entity to perform the following operations: Map the ML model configuration file to the identifier list of ML model tasks; and The third network entity is discovered based on the mapping.

5. The first network entity of claim 3, wherein the one or more criteria include the ML model task profile, wherein the at least one processor is configured to cause the first network entity to discover the third network entity based on the one or more criteria by causing the first network entity to perform the following operations: Map the ML model task configuration file to the identifier list of the pre-trained ML model; and The third network entity is discovered based on the mapping.

6. The first network entity according to any preceding claim, wherein the at least one processor is configured to cause the first network entity to obtain the first information from the third network entity by causing the first network entity to perform the following operations: A second request for an available pre-trained ML model is sent to the third network entity, wherein the second request includes the one or more criteria for identifying the at least one pre-trained ML model; and Receive a second response, including the first information, from the third network entity.

7. The first network entity according to any of the preceding claims, wherein the at least one processor is configured to enable the first network entity to evaluate the one or more available pre-trained ML models as pre-trained ML models for the ML model task.

8. The first network entity of claim 7, wherein the at least one processor is configured to cause the first network entity to evaluate the one or more available pre-trained ML models by causing the first network entity to perform the following operations: The one or more available pre-trained ML models are rated to generate rating parameters.

9. The first network entity of claim 8, wherein the at least one processor is configured to cause the first network entity to rate the one or more available pre-trained ML models based on one or more similarities to the ML model task, wherein the one or more similarities include similarities based on: One or more features of the pre-trained ML model; The training data requirements of the pre-trained ML model; The ML model algorithm of the pre-trained ML model; Previous use of the pre-trained ML model; The previous ratings of the pre-trained ML model; and The rating of the source of the pre-trained ML model.

10. The first network entity according to any one of claims 8 to 9, wherein the at least one processor is configured to enable the first network entity to rate the one or more available pre-trained ML models for the applicability of the following identifiers: analyze; ML modeling tasks; or ML model configuration file.

11. The first network entity according to any one of claims 8 to 10, wherein the second information includes the rating parameter.

12. The first network entity according to any of the preceding claims, wherein the second information includes the one or more available pre-trained ML models.

13. The first network entity according to any of the preceding claims, wherein: The first network entity includes an AIML (Artificial Intelligence Machine Learning) enabled server; The second network entity includes an application data analytics enabler server 'ADAES', an edge application server, or a vertical application layer 'VAL' server; and / or The third network entity includes a pre-trained model repository 'PMR' or a pre-trained model provider 'PMP', wherein optionally: The PMR is the application layer analytical data repository function 'AADRF', the general API framework core function 'CCF', the analytical data repository function 'ADRF', or the edge repository; The PMP is a VAL server, AIML enable server, or model training and management entity 'MTME'.

14. The first network entity according to any of the preceding claims, wherein the ML model task is an analysis task.

15. A second network entity for wireless communication, comprising: At least one memory; and At least one processor, coupled to the at least one memory and configured to enable the second network entity to: A first request to identify at least one pre-trained machine learning 'ML' model for transfer learning is sent to a first network entity, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; and Receive second information from the first network entity, the second information indicating one or more available pre-trained ML models as pre-trained ML models for the transfer learning of the ML model task.

16. The second network entity according to claim 15, wherein: The first network entity includes an AIML (Artificial Intelligence Machine Learning) enabled server; and / or The second network entity includes the application data analytics enabler server 'ADAES' or the vertical application layer 'VAL' server.

17. A third network entity for wireless communication, comprising: At least one memory; and At least one processor, coupled to the at least one memory and configured to enable the third network entity to: Receive a second request for an available pre-trained ML model from a first network entity, wherein the second request includes one or more criteria for identifying at least one pre-trained ML model; Based on the one or more criteria, first information for one or more available pre-trained ML models is determined; and A second response, including the first information of one or more available pre-trained ML models, is sent to the first network entity.

18. The third network entity according to claim 17, wherein: The first network entity includes an AIML (Artificial Intelligence Machine Learning) enabled server; and / or The third network entity includes a pre-trained model repository 'PMR' or a pre-trained model provider 'PMP', wherein optionally: The PMR is the application layer analytical data repository function 'AADRF', the core function of the general API framework 'CCF', or the edge repository; The PMP is a VAL server, AIML enable server, or model training and management entity 'MTME'.

19. A method performed by a first network entity, comprising: A first request is received from a second network entity to identify at least one pre-trained machine learning 'ML' model for transfer learning, the transfer learning being used for an ML model task, wherein the first request includes one or more criteria for identifying the at least one pre-trained ML model; Based on one or more of the criteria, a third network entity is discovered to indicate the availability of the pre-trained ML model; Based on the one or more criteria, first information for one or more usable pre-trained ML models is obtained from the third network entity; and The second information is provided to the second network entity, the second information indicating that the one or more available pre-trained ML models are used as pre-trained ML models for the transfer learning of the ML model task.

20. The method of claim 19, wherein: The first network entity includes an AIML (Artificial Intelligence Machine Learning) enabled server; The second network entity includes an application data analytics enabler server 'ADAES' or a vertical application layer 'VAL' server; and / or The third network entity includes a pre-trained model repository 'PMR' or a pre-trained model provider 'PMP', wherein optionally: The PMR is the application layer analytical data repository function 'AADRF', the core function of the general API framework 'CCF', or the edge repository; The PMP is a VAL server, AIML enable server, or model training and management entity 'MTME'.

Images