System and method to discover models for transfer learning

Abstract

The disclosure relates to a 5G or 6G communication system for supporting a higher data transmission rate. The present disclosure provides a method and system for discovering an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning. receiving, from a second entity, a first request message for requesting to store machine learning (ML) model information; verifying whether the second entity is authorized to store the ML model, wherein the ML model is identified by the analytics ID included in the first request message; storing the ML model information, based on the request message and a result of the verification for the second entity; and transmitting, to the second entity, a first response message, as a response of the request message, for indicating an identifier for the ML model.

Description

SYSTEM AND METHOD TO DISCOVER MODELS FOR TRANSFER LEARNING

[0001] The present disclosure relates to the field of Artificial Intelligence (AI) / Machine Learning (ML). Particularly, the present disclosure relates to system and method to discover AI / ML models for transfer learning.

[0002] 5G mobile communication technologies define broad frequency bands such that high transmission rates and new services are possible, and can be implemented not only in "Sub 6GHz" bands such as 3.5GHz, but also in "Above 6GHz" bands referred to as mmWave including 28GHz and 39GHz. In addition, it has been considered to implement 6G mobile communication technologies (referred to as Beyond 5G systems) in terahertz bands (for example, 95GHz to 3THz bands) in order to accomplish transmission rates fifty times faster than 5G mobile communication technologies and ultra-low latencies one-tenth of 5G mobile communication technologies.

[0003] At the beginning of the development of 5G mobile communication technologies, in order to support services and to satisfy performance requirements in connection with enhanced Mobile BroadBand (eMBB), Ultra Reliable Low Latency Communications (URLLC), and massive Machine-Type Communications (mMTC), there has been ongoing standardization regarding beamforming and massive MIMO for mitigating radio-wave path loss and increasing radio-wave transmission distances in mmWave, supporting numerologies (for example, operating multiple subcarrier spacings) for efficiently utilizing mmWave resources and dynamic operation of slot formats, initial access technologies for supporting multi-beam transmission and broadbands, definition and operation of BWP (BandWidth Part), new channel coding methods such as a LDPC (Low Density Parity Check) code for large amount of data transmission and a polar code for highly reliable transmission of control information, L2 pre-processing, and network slicing for providing a dedicated network specialized to a specific service.

[0004] Currently, there are ongoing discussions regarding improvement and performance enhancement of initial 5G mobile communication technologies in view of services to be supported by 5G mobile communication technologies, and there has been physical layer standardization regarding technologies such as V2X (Vehicle-to-everything) for aiding driving determination by autonomous vehicles based on information regarding positions and states of vehicles transmitted by the vehicles and for enhancing user convenience, NR-U (New Radio Unlicensed) aimed at system operations conforming to various regulation-related requirements in unlicensed bands, NR UE Power Saving, Non-Terrestrial Network (NTN) which is UE-satellite direct communication for providing coverage in an area in which communication with terrestrial networks is unavailable, and positioning.

[0005] Moreover, there has been ongoing standardization in air interface architecture / protocol regarding technologies such as Industrial Internet of Things (IIoT) for supporting new services through interworking and convergence with other industries, IAB (Integrated Access and Backhaul) for providing a node for network service area expansion by supporting a wireless backhaul link and an access link in an integrated manner, mobility enhancement including conditional handover and DAPS (Dual Active Protocol Stack) handover, and two-step random access for simplifying random access procedures (2-step RACH for NR). There also has been ongoing standardization in system architecture / service regarding a 5G baseline architecture (for example, service based architecture or service based interface) for combining Network Functions Virtualization (NFV) and Software-Defined Networking (SDN) technologies, and Mobile Edge Computing (MEC) for receiving services based on UE positions.

[0006] As 5G mobile communication systems are commercialized, connected devices that have been exponentially increasing will be connected to communication networks, and it is accordingly expected that enhanced functions and performances of 5G mobile communication systems and integrated operations of connected devices will be necessary. To this end, new research is scheduled in connection with eXtended Reality (XR) for efficiently supporting AR (Augmented Reality), VR (Virtual Reality), MR (Mixed Reality) and the like, 5G performance improvement and complexity reduction by utilizing Artificial Intelligence (AI) and Machine Learning (ML), AI service support, metaverse service support, and drone communication.

[0007] Furthermore, such development of 5G mobile communication systems will serve as a basis for developing not only new waveforms for providing coverage in terahertz bands of 6G mobile communication technologies, multi-antenna transmission technologies such as Full Dimensional MIMO (FD-MIMO), array antennas and large-scale antennas, metamaterial-based lenses and antennas for improving coverage of terahertz band signals, high-dimensional space multiplexing technology using OAM (Orbital Angular Momentum), and RIS (Reconfigurable Intelligent Surface), but also full-duplex technology for increasing frequency efficiency of 6G mobile communication technologies and improving system networks, AI-based communication technology for implementing system optimization by utilizing satellites and AI (Artificial Intelligence) from the design stage and internalizing end-to-end AI support functions, and next-generation distributed computing technology for implementing services at levels of complexity exceeding the limit of UE operation capability by utilizing ultra-high-performance communication and computing resources.

[0008] The present disclosure relates to providing a method in discovering and utilizing AI / ML models for transfer learning in various application domains in wireless communication systems.

[0009] One or more shortcomings discussed above are overcome, and additional advantages and features are provided by the present disclosure. Other embodiments and aspects of the disclosure are described in detail herein and are considered a part of the disclosure.

[0010] In an embodiment, a method for discovering an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning is disclosed. The method comprises receiving, from a second entity, a first request message for requesting to store machine learning (ML) model information; verifying whether the second entity is authorized to store the ML model, wherein the ML model is identified by the analytics ID included in the first request message; storing the ML model information, based on the request message and a result of the verification for the second entity; and transmitting, to the second entity, a first response message, as a response of the request message, for indicating an identifier for the ML model.

[0011] In another embodiment, a method for discovering an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning is disclosed. The method comprises: transmitting, to a first entity, a first request message for requesting to store machine learning (ML) model information; and in case that the second entity is authorized to store the ML model identified by the analytics ID included in the first request message, receiving, from the first entity, a first response message, as a response of the first request message, for indicating an identifier for the ML model.

[0012] In yet another embodiment, a method for discovering an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning is disclosed. The method comprises: transmitting, to a first entity, a second request message for requesting to discover ML model information; in case that the third entity is authorized to discover at least one ML model, receiving, from the first entity, a second response message, as a response of the second request message, including information on at least one appropriate ML model.

[0013] In yet another embodiment, a first entity to discover an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning is disclosed. The first entity comprises: memory; a transceiver configured to transmit and receive signals; and at least one processor coupled to the transceiver and configured to: receive, from a second entity, a first request message for requesting to store machine learning (ML) model information; verify whether the second entity is authorized to store the ML model, wherein the ML model is identified by the analytics ID included in the first request message; store the ML model information, based on the request message and a result of the verification for the second entity; and transmit, to the second entity, a first response message, as a response of the request message, for indicating an identifier for the ML model.

[0014] In yet another embodiment, a second entity to discover an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning is disclosed. The second entity comprises: memory; a transceiver configured to transmit and receive signals; and at least one processor coupled to the transceiver and configured to: transmit, to a first entity, a first request message for requesting to store machine learning (ML) model information; and in case that the second entity is authorized to store the ML model identified by the analytics ID included in the first request message, receive, from the first entity, a first response message, as a response of the first request message, for indicating an identifier for the ML model.

[0015] In yet another embodiment, a third entity to discover an Artificial Intelligence / Machine Learning (AI / ML) model for transfer learning is disclosed. The third entity comprises: memory; a transceiver configured to transmit and receive signals; and at least one processor coupled to the transceiver and configured to: transmit, to a first entity, a second request message for requesting to discover ML model information; in case that the third entity is authorized to discover at least one ML model, receive, from the first entity, a second response message, as a response of the second request message, including information on at least one appropriate ML model.

[0016] The foregoing solution is illustrative only and is not intended to be in any way limiting. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features will become apparent by reference to the drawings and the following detailed description.

[0017] The methods and systems provided by the present disclosure enable efficient discovery, management, and utilization of AI / ML models for transfer learning across various application domains based on relevant discovery criteria and operational requirements.

[0018] To easily identify the discussion of any particular element or act, the most significant digit(s) in a reference number refer(s) to the figure number in which that element is first introduced.

[0019] FIG. 1 illustrates an environment for transfer learning in accordance with an embodiment of the present disclosure.

[0020] FIG. 2 illustrates a sequence diagram for storing AI / ML models for transfer learning in accordance with an embodiment of the present disclosure.

[0021] FIG. 3 illustrates a sequence diagram for discovering the AI / ML models for transfer learning in accordance with an embodiment of the present disclosure.

[0022] FIG. 4 illustrates a block diagram of a repository server in accordance with an embodiment of the present disclosure.

[0023] FIG. 5 illustrates a flow diagram of a method for discovering the AI / ML models for transfer learning in accordance with an embodiment of the present disclosure.

[0024] It should be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of the　illustrative systems embodying the principles of the present subject matter. Similarly, it will be appreciated that any flowcharts, flow diagrams, state transition diagrams, pseudo code, and the like represent various processes which may be substantially represented in computer readable medium and executed by a computer or processor, whether or not such computer or processor is explicitly shown.

[0025] For the purpose of promoting an understanding of the principles of the invention, reference will now be made to the various embodiments and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, such alterations and further modifications in the illustrated system, and such further applications of the principles of the invention as illustrated therein being contemplated as would normally occur to one skilled in the art to which the invention relates.

[0026] It will be understood by those skilled in the art that the foregoing general description and the following detailed description are explanatory of the invention and are not intended to be restrictive thereof.

[0027] In the present document, the word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any embodiment or implementation of the present subject matter described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.

[0028] While the disclosure　is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will be described in detail below. It should be understood, however, that it is not intended to limit the disclosure to the particular forms disclosed, but on the contrary, the disclosure is to cover all modifications, equivalents, and alternative falling within the spirit and the scope of the disclosure.

[0029] The terms "comprise", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a setup, device, or method that comprises a list of components or steps　does not include only those components or steps but may include other components or steps not expressly listed or inherent to such setup or device or method. In other words, one or more elements in a device or system or apparatus proceeded by "comprises a" does not, without more constraints, preclude the existence of other elements or additional elements in the device or system or apparatus.

[0030] The terms like "at least one" and "one or more" may be used interchangeably throughout the description.　The terms like "AI / ML model" and "model" may be used interchangeably throughout the description. The terms like "repository server" and "server" may be used interchangeably throughout the description. The terms "authorizing entity", "storing entity" and "model owner" may be used interchangeably throughout the description. In the following detailed description of the embodiments of the disclosure, reference is made to the　accompanying drawings that form a part hereof, and in which are shown by way of illustration specific embodiments in which the disclosure may be practiced. These embodiments are described in sufficient detail to enable those skilled in art to practice the disclosure, and it is to be understood that other embodiments may be utilized and that changes may be made without departing from the scope of the present disclosure. The following description is, therefore, not to be taken in a limiting sense. In the following description, well known functions or constructions are not described in detail since they would obscure the description with unnecessary detail.

[0031] The present disclosure relates to an Artificial Intelligence / Machine Learning (AI / ML) models discovering system. The system takes into account certain discovery parameters comprised in a discovery request to discover the AI / ML models. The present disclosure discovers the AI / ML models for enabling transfer learning. As of now, only the concept of transfer learning is known but no method / system exists that can discover the AI / ML models to enable transfer learning. Additionally, the present disclosure provides a method / system for storing the AI / ML models at a server that can be discovered for transfer learning.

[0032] Artificial Intelligence (AI) / Machine Learning (ML) is being used in a range of application domains across industry sectors. In mobile communications systems, mobile devices (e.g. smartphones, automotive, robots) are increasingly replacing conventional algorithms (e.g. speech recognition, image recognition, video processing) with AIML models to enable applications.

[0033] In practical deployments involving AIML operations, the number of available models is not that large to cover all possible environmental conditions in a network. If existing model is used for the environment for which model was not trained with sufficient amount of data, then the model performance is reduced in terms of accuracy or precision during inference. On other hand, training the model with new environment requires considerable amount of training data and time. Sometimes the amount of required training data is also not available. Also, sometimes for faster convergence and performance, learning based on an existing learnt model helps.

[0034] Transfer learning (TL) is a technique in machine learning (ML) in which knowledge learned from a task is re-used in order to boost performance on a related task. In TL, the training of a model (to solve a particular task) is carried out using information from a "similar" domain in which enough information is available (the so-called source domain). With TL, some parts of the model trained in the source domain are used and further fine-tuned with the information available in the target domain. This way, the model reuses the understanding of the structure of the problem in similar domain and does not have to learn from beginning and thus resulting in faster model convergence / training.

[0035] In order to support transfer of learning, it is required to discover appropriate model from similar domain or environment. Further, consumer of the model also needs authorization (or consent from the model owner) in order to re-use the model.

[0036] In an embodiment of the present disclosure, entity-1 (model owner) stores an Artificial Intelligence / Machine Learning (AIML) model into entity-2 with application specific metadata, which includes but not limited to, type of application, sharing preference, observed inference accuracy, and observed latency for execution, current state of the model, and the like.

[0037] In an embodiment of the present disclosure, entity-3 requests to discover the AIML model with specific discovery criteria, which includes but not limited to, type of service, scope of the model, required accuracy performance, required latency performance, and the like.

[0038] In an embodiment of the present disclosure, the entity-2 determines candidate model by matching discovery filters along with AIML model metadata for which current state indicates that AIML model is trained.

[0039] In an embodiment of the present disclosure, the entity-2 decides model as discovered model based on sharing preference of the AIML model, and adds the model into discovered model list and shares the discovered models to the entity-3.

[0040] In 3rd Generation Partnership Project (3GPP) TR 23.700-82 (v0.2.0), AIML model storage procedure and discovery procedures are defined. However, existing procedures do not support discovery of model for transfer of learning for similar type of application or domain. Further, the existing procedures do not support authorizing the model information access, obtaining the authorization from the model owner for model information access.

[0041] The present disclosure facilitates enhancements that are needed to support transfer of learning. In an embodiment, the present disclosure describes the required enhancements to the AIML model information storage procedure. In step 1, entity-1 (e.g. ML Model owner or Application Data Analytics Enabler (ADAE) server or Vertical Application Layer (VAL) server or ADAE client or User Equipment (UE) or Network Exposure Function (NEF) or application server, like so) sends a request to entity-2 (e.g. Application layer - Analytical Data Repository Function (A-ADRF) or ADAE server or Application layer - Data Collection and Coordination Function (A-DCSF) or VAL server or ADAE client or UE or Network Exposure Function (NEF) or application server, like so) to store ML model information. The request may include, but not limited to, type of service or type of application or category of the application or domain (e.g. speech recognition, image recognition, video processing) in which the model is trained, sharing preference, observed inference accuracy and observed latency for execution along with other model specific information. Herein, the parameter sharing preference may include, but not limited to, the information about allowing access to model information.

[0042] In an embodiment, some of the possible values the parameter sharing preference may include:

[0043] a) "Consent required" - indicating that in order to use the model by the model discoverer / consumer / entity-2, an authorization from the entity-1 is required.

[0044] b) "Always allowed" - indicating that the model discoverer / consumer / entity-2 can use the model always without explicit consent from the entity-1.

[0045] c) "List of allowed consumers" - providing list of consumers who are allowed to use the model. The request further includes current state of the model (i.e. training in progress, trained, like so).

[0046] In an embodiment, the request includes an indication whether model is discoverable by other entities or not.

[0047] In step 2, the entity-2 authenticates and stores the model information.

[0048] In step 3, in the response message to the entity-1, the entity-2 provides model identifier and address of the storage where the model is stored.

[0049] In one embodiment, the present disclosure describes the required enhancements to the AIML model information discovery procedure. In step 1, entity-3 (e.g. ML Model consumer or ADAE server or VAL server or ADAE client or UE or NEF or application server or application client, like so) sends request to entity-2 (e.g. A-ADRF or ADAE server or A-DCSF or VAL server or ADAE client or UE or Network Exposure Function (NEF) or application server, like so) to discover ML model information. The request includes, but not limited to, type of service (e.g. image processing, location prediction, etc.) for which the model may be used, scope of the model (service scaling, or service arbitration), required accuracy performance, required latency performance.

[0050] In step 2, Upon receiving the request, the entity-2 may authenticate the entity-3 and discovers the appropriate list of AIML models as candidate models based on the state of the model whether model is trained or not.

[0051] In an embodiment, for each candidate model, based on sharing preference value of the model (as set by the entity-1), the entity-2 may decide the following actions:

[0052] - If the sharing preference is set to consent required (or any other equivalent value indicating consent from the owner is required), then based on policy, the entity-2 either performs steps 3-4 to obtain authorization from entity-1 or adds the model related information / model metadata into discovered model list and indicates the entity-3 to obtain authorization directly from entity-1.

[0053] - If the sharing preference is set to always allowed (or any other equivalent value indicating that the requesting consumer can always use the model), then the entity-2 may add the model related information / model metadata into discovered model list, and skips step-3-4).

[0054] - If the sharing preference value is set to list of allowed consumers, then the entity-2 may check whether the entity-3 is present in the allowed consumer list or not. If it is present in the allowed list then the entity-2 adds the model into the discovered model list. Otherwise, the entity-2 may not add the model related information / model metadata into the discovered model list, and skips step-3-4).

[0055] In steps 3&4, the entity-2 may send a request to entity-1 to authorize the entity-3 to use the model. The request includes, but not limited to, identity of the entity-3 and model identity. The entity-1 may either accepts or rejects the request. If the entity-1 has authorized the entity-3 to use the model in step-4, then the entity-2 may add the model into the discovered model list.

[0056] In step 5, the entity-2 may send a response to the entity-3 including the result of the request. If the entity-2 determines the discovered model list then it includes the list into the response along with model details. If the entity-3 requires to take explicit authorization from the entity-1, then the entity-2 indicates the same in the response and also provides URL address of the entity-1 where the entity-3 can contact entity-1 to request the authorization.

[0057] In an embodiment, sharing preference value may include direct consent required from model owner. When such a value is included in the sharing preference value, the entity-2 responds to entity-3 requesting the entity-3 to take explicit authorization from entity-1 and provides URL address of entity-1 to entity-3 to obtain authorization.

[0058] In an embodiment, the entity-1 is Network Data Analytics Function (NWDAF) service consumer or NWDAF containing Analytics Logical Lunction (AnLF) or AIML client or ADAES client or VAL client or ADAE server or AIML enablement server or VAL server or Network exposure function (NEF) or any other consumer client or any other analytics consumer.

[0059] In an embodiment, the Entity-3 is NWDAF service consumer or NWDAF containing AnLF or AIML client or ADAES client or VAL client or ADAE server or AIML enablement server or VAL server or Network exposure function (NEF) or any other consumer client or any other analytics consumer.

[0060] The following description with reference to the accompanying drawings is provided to assist in a comprehensive understanding of various embodiments of the present disclosure as defined by the claims and their equivalents. The following description includes various specific details to assist in that understanding but these are to be regarded as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the various embodiments described herein can be made without departing from the scope and spirit of the present disclosure. In addition, descriptions of well-known functions and constructions may be omitted for clarity and conciseness.

[0061] FIG. 1 illustrates an environment 100 for transfer learning in accordance with an embodiment of the present disclosure. In an embodiment, the environment 100 may comprise various elements such as a requesting entity 102, a repository server 104, and an authorizing entity 106, but not limited thereto. The requesting entity 102 may be an AI / ML model consumer, the repository server 104 may be an application server, and the authorizing entity 106 may be an AI / ML model owner.

[0062] In an embodiment, the requesting entity 102, the repository server 104, and the authorizing entity 106 may be any one among a User Equipment (UE), an Application layer- Analytical Data Repository Function (A-ADRF), an Application layer - Data Collection and Coordination Function (A-DCSF), an application server, an application client, a Network Data Analytics Function (NWDAF) service consumer, a NWDAF containing Analytics Logical Function (AnLF), an AI / ML client, an Application Data Analytics Enabler (ADAE) server, an ADAE server client, a Vertical Application Layer (VAL) client, a VAL server, an AI / ML enablement server, a Network Exposure Function (NEF) or any other consumer client or any other analytics consumer, but not limited thereto.

[0063] In an embodiment, the requesting entity 102 may communicate with the repository server 104 over a communication network and the repository server 104 may communicate with the authorizing entity 106 over another communication network. In another embodiment, the requesting entity 102 may directly communicate with the authorizing entity 106. The requesting entity 102 may request the repository server 104 to discover AI / ML models for transfer learning and the authorizing entity 106 may authorize the requesting entity 102 to access the AI / ML models for the transfer learning. The functionality of the requesting entity 102, the repository server 104, and the authorizing entity 106 are explained in more detail in the upcoming paragraphs.

[0064] FIG. 2 illustrates a sequence diagram for storing AI / ML models for transfer learning in accordance with an embodiment of the present disclosure. In an embodiment, storing entity 202 of FIG. 2 may be the same as the authorizing entity 106 of FIG.1.

[0065] At step 204, the storing entity 202 may send a storing request i.e., AI / ML model information storage request to the repository server 104 to store an AI / ML model information. The storing entity 202 may be the owner of the AI / ML model. The storing request may include one or more information elements as described in Table-1 below.

[0066] Information elementStatusDescriptionRequester IdentityMThe identity of the model owner or AI / ML repository consumer performing the request.AI / ML modelO(NOTE　1)Provides the trained AI / ML model.AI / ML model informationO(NOTE　1)Provides information of the AI / ML model, as described in Table 2 below.NOTE　1: At least one of these information elements shall be provided.

[0067] The status 'M' may indicate a mandatory information element, and the status 'O' may indicate an optional information element.

[0068] Information elementStatusDescriptionAI / ML model identifierMAn identifier for the AI / ML modelAI / ML model addressORepresents the AI / ML model address that can be used by repository server to download the AI / ML model.Analytics IDORepresents analytics ID for which the model can be used.VAL service ID(s)OIdentify the VAL service ID(s).DomainOSpecifies domain for which the model can be used (e.g., for speech recognition, image recognition, video processing, location predication, etc.).Sharing preferenceO(NOTE　1)Indicates information about allowing access to the model information stored at the repository server.AI / ML model interoperability informationO(NOTE　1)Represents the model vendor-specific information that conveys, e.g., requested model file format, model execution environment, etc. The encoding, format, and value of AI / ML Model Interoperable Information is not specified since it is vendor specific information, and is agreed between vendors, if necessary for sharing purposes.AI / ML Model phaseO(NOTE　1)Represents the AI / ML model phase, e.g., in training, trained, re-training, deployed.> Observed interference accuracyO(NOTE 2)Provides accuracy of the model (if AI / ML model is in trained phase).AI / ML model storage and discovery requirementsO(NOTE　1)Represents the requirements for the AI / ML repository for the AI / ML model storage and discovery.> Storage durationORepresents the AI / ML model storage duration time. When the storage duration time is expired, the stored AI / ML model and the related information shall be deleted.> Security and access requirementsORepresents the information on security requirements for storing the AI / ML model information and the AI / ML model access requirements (e.g., publicly available, private use only, or available for the list of VAL server IDs).NOTE　1: At least one of these information elements shall be provided.NOTE　2: This information element is included only if trained ML model is available.

[0069] In an embodiment, the information element sharing preference may have a value as "consent required" indicating that in order to use the model by a model consumer, an authorization is required from the model owner. In an embodiment, the information element sharing preference may have a value as "always allowed" indicating that in order to use the model, the model consumer do not require any explicit authorization from the model owner. In an embodiment, the model owner may provide a "list of allowed model consumers" under the information element sharing preference indicating the model consumers that are allowed to retrieve and use the model.

[0070] Additionally, the storing request may also include an indication whether the model is discoverable by the model consumer or any other entity, or not.

[0071] In an embodiment, the model interoperability information may also include observed latency for execution of the model along with other information like file format and model execution environment.

[0072] At step 206, the repository server 104 may authenticate if the storing entity 202 is authorized to store the AI / ML model identified by the analytics ID and / or list of allowed model consumers within the AI / ML model profile attribute. If the storing entity 202 is authorized, the repository server 104 may process the storing request and store the information of the AI / ML model.

[0073] At step 208, the repository server 104 may send a confirmation response to the storing entity 202 with an identifier of the created AI / ML model profile as described in Table-3 below.

[0074] Information elementStatusDescriptionResultMIndicates success or failure of the request.AI / ML model addressORepresents the AI / ML model address that can be used by consumer to download the model. This information element (IE) is included when result IE indicates success.AI / ML model identifierMAn identifier for the ML model. This IE is included when result IE indicates success.

[0075] FIG. 3 illustrates a sequence diagram for discovering the AI / ML models for transfer learning in accordance with an embodiment of the present disclosure. At step 302, the requesting entity 102 may send a discovery request i.e., AI / ML model information discovery request to the repository server 104 to discover AI / ML models information. The discovery request may include one or more information elements or discovery parameters (as described in Table- 4 below) for discovering the AI / ML models information.

[0076] Information elementStatusDescriptionRequester IdentityMThe identity of the AI / ML repository consumer performing the request.Filtering criteriaMRepresents the filtering criteria, which can be any of the AI / ML model information as in Table-1.> DomainMDomain of the service in which the model will be used to transfer learning (e.g., image processing, location predication, etc.)> Required accuracy performanceORequired accuracy for the ML model (e.g. for transfer learning).> Analytics IDO(NOTE　1)Represents analytics ID for which the model can be used.> Training LevelOIndicates required training level of the AI / ML model (if the model is in training phase).> Required ML model interoperabilityO(NOTE　1)Represents the vendor-specific information that conveys, e.g., requested model file format, model execution environment, etc. The encoding, format, and value of ML Model Interoperable Information is not specified since it is vendor specific information, and is agreed between vendors, if necessary for sharing purposes.> ML Model phaseO(NOTE 1)Required ML model phase, e.g., in training, trained, re-training, deployed.NOTE　1: At least one of these information elements shall be provided.

[0077] At step 304, upon receiving the discovery request, the repository server 104 may authenticate the identity of the requesting entity 102 to determine whether the requesting entity 102 is authorized to discover the AI / ML models or not. Further, the repository server 104 may map the one or more discovery parameters with the information elements of one or more stored AI / ML models at the repository server 104 and discover a list of one or more AI / ML models as candidate models based on mapping results of the one or more discovery parameters with the information elements of the stored AI / ML models.

[0078] In an embodiment, the domain and training phase of the model as described in the discovery parameters may need to be exactly mapped with the domain and training phase of the model as described by the information elements of the model. In an embodiment, discovery parameters other than the domain and training phase of the model may be approximately mapped with the information elements up to a threshold value described by the requesting entity 102.

[0079] In an embodiment, if for any model out of the one or more candidate models, the sharing preference is set to a list of allowed consumers by the model owner, then the repository server 104 may check whether the requesting entity 102 is present in the list of allowed consumers or not. If the requesting entity 102 is present in the list of allowed consumers, then the repository server 104 may add the candidate model related information into the discovered list of the models. However, if the requesting entity 102 is not present in the list of allowed consumers, then the repository server 104 may not add the candidate model related information into the discovered list of the models and may skip steps 306 and 308.

[0080] In another embodiment, if for any model out of the one or more candidate models, the sharing preference is set to consent required by the model owner then, at step 306, the repository server 104 may send an authorization request to the authorizing entity 106 for authorization of the requesting entity 102. The authorization request may include, but not limited to, identity of the requesting entity and identity of the candidate model. The authorizing entity 106 may either accept or reject the authorization request sent by the repository server 104.

[0081] At step 308, the authorizing entity 106 may send an authorization response to the repository server 104. If the authorizing entity 106 has authorized the requesting entity 102 to use the candidate model, then the repository server 104 may add the candidate model into the discovered list of the models. However, if the authorizing entity has not authorized the requesting entity 102 to use the candidate model, then the repository server 104 may not add the candidate model into the discovered list of the models.

[0082] At step 310, the repository server 104 may send a response including results of the discovery request to the requesting entity 102. If the repository server 104 determines the list of discovered models, then the repository server 104 may include the list into the response along with details of the models as described in Table-5 below.

[0083] Information elementStatusDescriptionResultMIndicates success or failure of the requestML model ID listORepresents the identifiers of the discovered list of AI / ML models. This information element shall be provided if the result is success.> ML modelO(NOTE　1)(NOTE　2)Represents the AI / ML model(s). This information element may be provided if the result is success.> ML model address(es)O(NOTE　1)(NOTE　2)Represents the AI / ML model address(es) at Model repository that can be used to download the AI / ML model. This information element may be provided if the result is success.NOTE　1: Only one of these information elements shall be provided.NOTE　2: For transfer learning, the list of all possible models matching the filtering criteria can be included along with model address.

[0084] FIG. 4 illustrates a block diagram of the repository server 104 in accordance with an embodiment of the present disclosure. The repository server 104 may include one or more components such as an Input / Output (I / O) interface 402, a processor 404 and a memory 406, but not limited thereto. The memory 406 may include one or more modules 408 such as a receiving module 410, a discovery module 412, and a sending module 414, but not limited thereto. It will be appreciated that such aforementioned modules may be represented as a single module or a combination of different modules. The modules may be implemented in any suitable hardware, software, firmware, or combination thereof. Further the modules may be implemented by various techniques comprising but not limited to computer programs, one or more neural networks, machine learning algorithms, embedded systems design and cloud computing architectures.

[0085] The I / O interface 402 may be in communication with the requesting entity 102 and the authorizing entity 106 to receive one or more requests and send one or more responses. The I / O interface 402 may employ communication protocols / methods such as, without limitation, audio, analog, digital, stereo, IEEE-1394, serial bus, Universal Serial Bus (USB), infrared, PS / 2, BNC, coaxial, component, composite, Digital Visual Interface (DVI), high-definition multimedia interface (HDMI), Radio Frequency (RF) antennas, S-Video, Video Graphics Array (VGA), IEEE 802.n / b / g / n / x, Bluetooth, cellular (e.g., Code-Division Multiple Access (CDMA), High-Speed Packet Access (HSPA+), Global System For Mobile Communications (GSM), Long-Term Evolution (LTE) or the like), etc. The memory 406 may store one or more instructions for the operation of one or more functions of the repository server 104. In an embodiment, the instructions stored in the memory 406 are processed by the modules 408 of the repository server 104. The modules 408 may be stored within the memory 406 as shown in FIG.4. As an example, the modules 406, communicatively coupled to the processing unit 404, may also be present outside the memory 406.The processor 404 may execute the one or more instructions stored in the memory 406 to perform the one or more functions of the repository server 104.

[0086] In an embodiment, the receiving module 410 may receive the storing request from the storing entity 202 via the I / O interface 402 to store the model at the repository server 104. The receiving module may also receive the discovery request from the requesting entity 102 via the I / O interface 402 to discover the models for enabling the transfer learning. The discovery module 412 may perform model discovery based on the discovery request received from the requesting entity 102. The sending module 414 may send the confirmation response to the storing entity 202 after storing the model via the I / O interface 402. The sending module 414 may also send the response including the list of the discovered models to the requesting entity 102 via the I / O interface 402.

[0087] In an embodiment, the repository server 104 may receive from the storing entity 202 using the receiving module 410, the storing request i.e., AI / ML model information storage request to store the AI / ML model information. The storing request may include the one or more information elements as described in Table-1, and the same has not been repeated for the sake of brevity. The repository server 104 may authenticate if the storing entity 202 is authorized to store the AI / ML model based on the one or more information elements included in the storing request. If the storing entity 202 is authorized, the repository server 104 may process the storing request and store the information of the AI / ML model. Upon storing the AI / ML model, the repository server 104 may send to the storing entity 202 using the sending module 414, a confirmation response with the identifier of the created AI / ML model profile as described in Table-3.

[0088] In an embodiment, the repository server 104 may receive from the requesting entity 102 using the receiving module 410, the discovery request i.e., AI / ML model information discovery request for discovering AI / ML models information. The discovery request may include one or more information elements or discovery parameters as described in Table- 4 for discovering the AI / ML models information. Upon receiving the discovery request, the repository server 104 may authenticate the identity of the requesting entity 102 to determine whether the requesting entity 102 is authorized to discover the AI / ML models or not. Further, the repository server 104 may map the one or more discovery parameters with the information elements of one or more stored AI / ML models at the repository server 104 and discover, using the discovery module 412, the list of the one or more AI / ML models as the candidate models based on the mapping results of the one or more discovery parameters with the information elements of the stored AI / ML models.

[0089] The repository server 104 may determine the sharing preference of the candidate models set by the model owners and based on the sharing preference of the candidate models, the repository server 104 may send the authorization request to the authorizing entity 106 for the authorization of the requesting entity 102 to access the candidate model. In response to the authorization request, the repository server 104 may receive the authorization response from the authorizing entity 106. If the authorizing entity 106 has authorized the requesting entity 102 to use the candidate model, then the repository server 104 may add the candidate model into the discovered list of the models. However, if the authorizing entity has not authorized the requesting entity 102 to use the candidate model, then the repository server 104 may not add the candidate model into the discovered list of the models. The repository server 104 may send the response including the results of the discovery request to the requesting entity 102. If the repository server 104 determines the list of discovered models, then the repository server 104 may include the list into the response along with the details of the models as described in Table-5, the same has not been repeated for the sake of brevity.

[0090] In an embodiment, for a candidate model, if the requesting entity 102 requires taking explicit authorization from the authorizing entity 106, then the repository server 104 may indicate the same in the response. The repository server 104 may provide Uniform Resource Locator (URL) address of the authorizing entity 106 where the requesting entity 102 may contact the authorizing entity 106 to request the authorization.

[0091] FIG. 5 illustrates a flow diagram 500 of a method for discovering the AI / ML models for transfer learning in accordance with an embodiment of the present disclosure. The blocks of the flow diagram shown in FIG. 5 have been arranged in a generally sequential manner for ease of explanation, however, it is to be understood that this arrangement is merely　exemplary, and it should be recognized that the functionality / processing associated with method 500 (and the blocks shown in FIG. 5) can occur in a different order (for example, where at least some of the functionality / processing associated with the blocks is performed in parallel and / or in an event-driven manner).

[0092] At step 502, the method comprises receiving a discovery request for the AI / ML model. The discovery request comprises one or more discovery parameters indicating discovery criteria for the AI / ML model. The discovery request may include one or more information elements or discovery parameters as described in the Table-4 for discovering the AI / ML models information. The requesting entity 102 may be authenticated to determine whether the requesting entity 102 is authorized to discover the AI / ML models or not.

[0093] At step 504, the method comprises discovering at least one AI / ML model, among a plurality of AI / ML models stored at the repository server, based on the one or more discovery parameters. The one or more discovery parameters may be mapped with the information elements of the one or more stored AI / ML models at the repository server 104 for discovering the AI / ML models. The plurality of AI / ML models are stored at the repository server by one or more authorizing entities. Storing the plurality of AI / ML models comprises receiving, by the repository server from the one or more authorizing entities, one or more storing requests to store the plurality of AI / ML models on the repository server, authorizing, by the repository server, the one or more authorizing entities to store the plurality of AI / ML models and sending, by the repository server to the one or more authorizing entities, one or more confirmation responses indicating successful authorization to store the plurality of AI / ML models upon successful authorization. In an embodiment, the method further comprises enabling, by the repository server, the requesting entity to send direct authorization request message to the at least one authorizing entity.

[0094] In yet another embodiment, the one or more storing requests comprise domain of the service in which the AI / ML model is trained, sharing preference of the AI / ML model including the list of allowed requesting entities that are allowed to retrieve and use the AI / ML model, observed interference accuracy of the AI / ML model and observed latency for execution of the AI / ML model along with other model specific information. The other model specific information comprises at least one of model file format and model execution environment.

[0095] The at least one AI / ML model may be discovered based on the mapping results of the one or more discovery parameters with the information elements of the stored AI / ML models. The sharing preference of the at least one discovered AI / ML model may be determined and based on the sharing preference of the at least one discovered AI / ML model, the authorization request may be sent to the authorizing entity 106 for the authorization of the requesting entity 102 to access the at least one discovered AI / ML model. In response to the authorization request, the authorization response may be received from the authorizing entity 106. If the authorizing entity 106 has authorized the requesting entity 102 to use the at least one discovered AI / ML model, then the at least one discovered AI / ML model may be added into the discovered list of the models. However, if the authorizing entity has not authorized the requesting entity 102 to use the at least one discovered AI / ML model, then the at least one discovered AI / ML model may not be added into the discovered list of the models. In an embodiment, the information about the at least one discovered AI / ML model comprises identifier of the AI / ML model and address of the AI / ML model, at the repository server, to access the AI / ML model.

[0096] At step 506, the method further comprises transmitting a response including information about the at least one discovered AI / ML model to enable transfer learning. If the list of discovered models is determined, then the list may be included in the response.

[0097] In this manner, the present disclosure may discover the AI / ML models stored at the repository server for enabling the transfer learning.

[0098] While　various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art.　The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the detailed description.

[0099] The order in which the various operations of the methods are described is not intended to be construed as a limitation, and any number of the described method blocks can be combined in any order to implement the method. Additionally, individual blocks may be deleted from the methods without departing from the spirit and scope of the subject matter described herein. Furthermore, the methods can be implemented in any suitable hardware, software, firmware, or combination thereof.

[0100] It may be noted here that the subject matter of some or all embodiments described with reference to FIG. 1-4 may be relevant for the methods and the same is not repeated for the sake of brevity. The various operations of methods described above may be performed by any suitable means capable of performing the corresponding functions. The means may include various hardware and / or software component(s) and / or module(s), including, but not limited to a circuit, an application specific integrated circuit (ASIC), or processor. Generally, where there are operations illustrated in Figures, those operations may be performed by any suitable corresponding counterpart means-plus-function components.

[0101] Furthermore, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term "computer-readable medium" should be understood to include tangible items and exclude carrier waves and transient signals, i.e., non-transitory. Examples include Random Access Memory (RAM), Read-Only Memory (ROM), volatile memory, nonvolatile memory, hard drives, Compact Disc (CD) ROMs, Digital Video Disc (DVDs), flash drives, disks, and any other known physical storage media.

[0102] Certain aspects may comprise a computer program product for performing the operations presented herein. For example, such a computer program product may comprise a computer readable media having instructions stored (and / or encoded) thereon, the instructions being executable by one or more processors to perform the operations described herein. For certain aspects, the computer program product may include packaging material.

[0103] Various components, modules, or units are described in this disclosure to emphasize functional aspects of devices configured to perform the disclosed techniques, but do not necessarily require realization by different hardware units. Rather, as described above, various units may be combined in a hardware unit or provided by a collection of interoperative hardware units, including one or more processors as described above, in conjunction with suitable software and / or firmware.

[0104] As used herein, a phrase referring to "at least one" or "one or more" of a list of items refers to any combination of those items, including single members. As an example, "at least one of: a, b, or c" is intended to cover: a, b, c, a-b, a-c, b-c, and a-b-c. The terms "a", "an" and "the" mean "one or more", unless expressly specified otherwise. The terms "including", "comprising", "having" and variations thereof, when used in a claim, is used in a non-exclusive sense that is not intended to exclude the presence of other elements or steps in a claimed structure or method, unless expressly specified otherwise.

[0105] Finally, the language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the invention be limited not by this detailed description, but rather by any claims that issue on an application based here on. Accordingly, the embodiments of the present disclosure are intended to be illustrative, but not limiting, of the scope of the invention, which is set forth in the appended claims.

Claims

1.A method performed by a first entity in a wireless communication system, the method comprising:receiving, from a second entity, a first request message for requesting to store machine learning (ML) model information;verifying whether the second entity is authorized to store the ML model, wherein the ML model is identified by the analytics ID included in the first request message;storing the ML model information, based on the request message and a result of the verification for the second entity; andtransmitting, to the second entity, a first response message, as a response of the request message, for indicating an identifier for the ML model.2.The method of claim 1, further comprising:receiving, from a third entity, a second request message for requesting to discover ML model information;verifying whether the third entity is authorized to discover at least one ML model;discovering an at least one appropriate ML model from the stored ML information, based on the second request message and a result of the verification for the third entity; andtransmitting, to the third entity, a second response message, as a response of the second request message, including information on the at least one appropriate ML model.3.The method of claim 2, wherein the first request message includes information on a current state of the ML model for indicating a current training state of the ML model, andwherein the at least one appropriate ML model is determined based on the information on the current state of the ML model.4.The method of claim 2, wherein the first request message includes information on a list of allowed entity for indicating at least one entity which is allowed to use the ML model, andwherein the at least one appropriate ML model is determined based on the information on the list of the allowed entity.5.A method performed by a second entity in a wireless communication system, the method comprising:transmitting, to a first entity, a first request message for requesting to store machine learning (ML) model information; andin case that the second entity is authorized to store the ML model identified by the analytics ID included in the first request message, receiving, from the first entity, a first response message, as a response of the first request message, for indicating an identifier for the ML model.6.The method of claim 5, wherein the first request message includes information on a current state of the ML model for indicating a current training state of the ML model and information on a list of allowed entity for indicating at least one entity which is allowed to use the ML model.7.A method performed by a third entity in a wireless communication system, the method comprising:transmitting, to a first entity, a second request message for requesting to discover ML model information;in case that the third entity is authorized to discover at least one ML model, receiving, from the first entity, a second response message, as a response of the second request message, including information on at least one appropriate ML model.8.The method of claim 7, wherein the at least one appropriate ML model is determined based on information on a current state of ML model for indicating a current training state of the ML model and information on a list of allowed entity for indicating at least one entity which is allowed to use the ML model.9.A first entity in a wireless communication system, comprising:memory;a transceiver configured to transmit and receive signals; andat least one processor coupled to the transceiver and configured to:receive, from a second entity, a first request message for requesting to store machine learning (ML) model information;verify whether the second entity is authorized to store the ML model, wherein the ML model is identified by the analytics ID included in the first request message;store the ML model information, based on the request message and a result of the verification for the second entity; andtransmit, to the second entity, a first response message, as a response of the request message, for indicating an identifier for the ML model.10.The first entity of claim 9, wherein the at least one processor is further configured to:receive, from a third entity, a second request message for requesting to discover ML model information;verify whether the third entity is authorized to discover at least one ML model;discover an at least one appropriate ML model from the stored ML information, based on the second request message and a result of the verification for the third entity; andtransmit, to the third entity, a second response message, as a response of the second request message, including information on the at least one appropriate ML model.11.The first entity of claim 10, wherein the first request message includes information on a current state of the ML model for indicating a current training state of the ML model, andwherein the at least one appropriate ML model is determined based on the information on the current state of the ML model.12.A second entity in a wireless communication system, comprising:memory;a transceiver configured to transmit and receive signals; andat least one processor coupled to the transceiver and configured to:transmit, to a first entity, a first request message for requesting to store machine learning (ML) model information; andin case that the second entity is authorized to store the ML model identified by the analytics ID included in the first request message, receive, from the first entity, a first response message, as a response of the first request message, for indicating an identifier for the ML model.13.The second entity of claim 12, wherein the first request message includes information on a current state of the ML model for indicating a current training state of the ML model and information on a list of allowed entity for indicating at least one entity which is allowed to use the ML model.14.A third entity in a wireless communication system, comprising:memory;a transceiver configured to transmit and receive signals; andat least one processor coupled to the transceiver and configured to:transmit, to a first entity, a second request message for requesting to discover ML model information;in case that the third entity is authorized to discover at least one ML model, receive, from the first entity, a second response message, as a response of the second request message, including information on at least one appropriate ML model.15.The third entity of claim 14, wherein the at least one appropriate ML model is determined based on information on a current state of ML model for indicating a current training state of the ML model and information on a list of allowed entity for indicating at least one entity which is allowed to use the ML model.

Images