Methods, apparatuses, computer programs and / or systems for node selection

Abstract

Node selection, such as node selection for federated learning based on for example receiving federated learning, FL, task information, generating a distributed nodes list indicating distributed nodes for contributing to a requested FL training, and transmitting the distributed nodes list, and updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list, wherein generating the distributed nodes list is based on a comparison of the characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes.

Description

TECHNICAL FIELD Various example embodiments of this disclosure relate to one or more methods, apparatuses, computer programs and / or systems for node selection, and more specifically (but not exclusively) to node selection in federated learning. BACKGROUND A communication system can be seen as a facility that enables communication sessions between two or more entities, such as communication devices, base stations and / or other nodes by providing carriers between the various entities involved in the communications path. The communication system may be a wireless communication system. Examples of wireless systems comprise PLMNs operating based on radio standards (such as, those provided by 3GPP), satellite based communication systems and different wireless local networks. A wireless local network may, for example, be implemented as a WLAN. The wireless systems can typically be divided into cells, and are therefore often referred to as cellular systems. The communication system and associated devices typically operate in accordance with a given standard or specification, which sets forth what the various entities associated with the system are permitted to do and how that should be achieved. Communication protocols and / or parameters, which shall be used for the connection, are also typically defined. An example of a set of standards includes the so-called 5G standards. SUMMARY Various example embodiments of this disclosure aim at addressing at least part of the issue and / or problems and drawbacks either explicitly described herein or otherwise apparent to a person skilled in the relevant art(s) to provide methods, apparatuses, computer programs, and / or systems by which mechanisms and / or procedures for node selection, such as node selection for federated learning, can be improved. Various example embodiments will be described with respect to certain aspects. These aspects are not intended to indicate key or essential features of the various example embodiments, nor are they intended to be used to otherwise limit the scope of this disclosure. Other features, aspects and elements of the various example embodiments will be readily apparent to a person skilled in the art in view of the disclosure. According to a first aspect, there is provided an apparatus (100'), comprising: means for receiving (150) federated learning, FL, task information including characteristics of a requested FL training, and means for generating (170), in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, a set of distributed nodes for contributing to the requested FL training, and means for transmitting (160) the distributed nodes list, wherein generating the distributed nodes list is based on a comparison of characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and based on stored availability information corresponding to the plurality of distributed nodes, and the apparatus further comprises storing means (180) configured for, in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list. According to an example aspect, the means for receiving (150) is further configured for receiving affiliation information comprising a list of affiliation nodes, indicating a subset of the set of distributed nodes indicated in the distributed nodes list, the storing means (180) is further configured for storing an affiliation entry including an identifier of the affiliation entry and the list of affiliation nodes, the means for receiving (150) is further configured for receiving second FL task information including characteristics of a second requested FL training, the means for generating (170) is further configured for generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and the means for transmitting (160) is further configured for transmitting the second distributed nodes list, and wherein generating the second distributed nodes list is further based on the stored affiliation entry. According to an example aspect, the means for transmitting (160) is further configured for transmitting the identifier of the affiliation entry, the means for receiving (150) is further configured for receiving affiliation update information including an identifier of a corresponding affiliation entry, and the storing means (180) is further configured for, in response to receiving the affiliation update information, updating the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation update information. According to an example aspect, the means for transmitting (160) is further configured for transmitting the identifier of the affiliation entry, the means for receiving (150) is further configured for receiving affiliation delete information including an identifier of a corresponding affiliation entry, and the storing means (180) is further configured for, in response to receiving the affiliation delete information, deleting the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation delete information. According to an example aspect, the storing means (180) is further configured for updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list in response to receiving the affiliation information. According to an example aspect, the storing means (180) is further configured for updating the stored availability information corresponding to the affiliation nodes in response to receiving the affiliation delete information and / or in response to receiving the affiliation update information. According to an example aspect, the stored availability information corresponding to the plurality of distributed nodes comprises at least one of a status of the plurality of distributed nodes, or one or more affiliation entries, each indicating a set of distributed nodes selected for a running FL training process. According to an example aspect, the storing means (180) is further configured for storing a list indicating reserved resources of distributed nodes in response to generating the distributed nodes list, the reserved resources of distributed nodes being reserved for the requested FL training; the means for receiving (150) is further configured for receiving second FL task information including characteristics of a second requested FL training, generating, in response to receiving the second FL task information, the second distributed nodes list is further based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes, and the means for transmitting (160) is further configured for transmitting the second distributed nodes list, wherein resources of distributed nodes indicated in the list indicating reserved resources of distributed nodes are excluded from the second distributed nodes list. According to an example aspect, the received affiliation information further comprises a predicted FL task completion time, the storing means (180) is configured for storing the affiliation entry such that the affiliation entry further comprises storing the predicted FL task completion time, and wherein the second distributed nodes list (S35) indicates a set of currently available distributed nodes and further indicates a set of currently unavailable distributed nodes together with an indication of a future time, when the currently unavailable distributed nodes will be available. According to an example aspect, the received affiliation information further comprises a first event exposure configuration, the second FL task information further comprises a second event exposure configuration, the means for generating is further configured for evaluating the first event exposure configuration and the second event exposure configuration, and the means for generating is further configured for, in case the first event exposure configuration and the second event exposure configuration match, transmitting an event report and generating the second distributed nodes list only after the FL training process corresponding to the first event exposure configuration is finished. According to an example aspect, the received affiliation information further comprises FL process information about an FL training process, the storing means (180) is configured for storing the affiliation entry such that the affiliation entry further comprises storing the FL process information and an identifier of a corresponding FL aggregator, the means for receiving (150) is further configured for receiving affiliation retrieve information or second FL task information, the received information including second FL process information, and the apparatus further comprises means for searching for an affiliation entry with FL process information that fits the second FL process information of the received information, and wherein the means for transmitting (160) is further configured for transmitting an affiliation retrieve response comprising the identifier of a corresponding FL aggregator included in the affiliation entry with FL process information that fits the second FL process information of the received information. According to an example aspect, the storing means (180) is further configured for updating the stored availability information based on a timer and / or periodically. According to a second aspect, there is provided a method, comprising: receiving federated learning, FL, task information including characteristics of a requested FL training (SI), generating, in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, distributed nodes for contributing to the requested FL training (S2), and transmitting the distributed nodes list (S4), wherein generating the distributed nodes list is based on a comparison of the characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and the method further comprising in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list (S3). According to another example aspect, the method further comprises receiving affiliation information comprising a list of affiliation nodes, indicating a subset of the set of distributed nodes indicated in the distributed nodes list (S6), creating an affiliation entry including an identifier of the affiliation entry and the list of affiliation nodes, storing the affiliation entry including the identifier of the affiliation entry and the list of affiliation nodes, receiving second FL task information including characteristics of a second requested FL training, generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and transmitting the second distributed nodes list, wherein generating the second distributed nodes list is further based on the stored affiliation entry. According to another example aspect, the method further comprises transmitting the identifier of the affiliation entry, receiving affiliation update information including an identifier of a corresponding affiliation entry, and in response to receiving the affiliation update information, updating the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation update Information. According to another example aspect, the method further comprises transmitting the identifier of the affiliation entry, receiving affiliation delete information including an identifier of a corresponding affiliation entry, and in response to receiving the affiliation delete information, deleting the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation delete information. According to another example aspect, the method further comprises updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list in response to receiving the affiliation information. According to another example aspect, the method further comprises updating (S17) the stored availability information corresponding to the affiliation nodes in response to receiving the affiliation delete information and / or in response to receiving the affiliation update information. According to another example aspect, the stored availability information corresponding to the plurality of distributed nodes comprises at least one of a status of the plurality of distributed node, or one or more affiliation entries, each indicating a set of distributed nodes selected for a running FL training process. According to another example aspect, the method further comprises storing a list indicating reserved resources of distributed nodes in response to generating the distributed nodes list; receiving second FL task information including characteristics of a second requested FL training, generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes, and transmitting the second distributed nodes list, wherein resources of distributed nodes indicated in the list indicating reserved resources of distributed nodes are excluded from the second distributed nodes list. According to another example aspect, the affiliation information further comprises a predicted FL task completion time, storing the affiliation entry further comprises storing the predicted FL task completion time, and the second distributed nodes list (S35) indicates a set of currently available distributed nodes and further indicates a set of currently unavailable distributed nodes together with an indication of a future time, when the currently unavailable distributed nodes will be available. According to another example aspect, the affiliation information further comprises a first event exposure configuration, the second FL task information further comprises a second event exposure configuration, and the method further comprises evaluating the first event exposure configuration and the second event exposure configuration, and in case the first event exposure configuration and the second event exposure configuration match, transmitting an event report and generating the second distributed nodes list only after the FL training process corresponding to the first event exposure configuration is finished. According to another example aspect, the affiliation information further comprises FL process information about an FL training, storing the affiliation entry further comprises storing the FL process information and an identifier of a corresponding FL aggregator, and the method further comprises receiving affiliation retrieve information or second FL task information, the received information including second FL process information, searching for an affiliation entry with FL process information that fits the second FL process information of the received information, and transmitting an affiliation retrieve response comprising the identifier of a corresponding FL aggregator included in the affiliation entry with FL process information that fits the second FL process information of the received information. According to another example aspect, the method further comprises updating the stored availability information based on a timer and / or periodically. According to a third aspect, there Is provided an apparatus (200'), comprising means for transmitting (260) federated learning, FL, task information including characteristics of a requested FL training, means for receiving (250) a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training, means for selecting (270) a subset of the set of distributed nodes, indicated by the received distributed nodes list, to be used for the requested FL training, wherein the means for transmitting (260) is further configured for transmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes. According to another example aspect, the means for receiving (250) is further configured for receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, the means for transmitting (260) is further configured for transmitting, in response to the requested FL training being executed and completed, affiliation delete information including the received identifier of the affiliation entry. According to another example aspect, the means for receiving (250) is further configured for receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, and the means for transmitting (260) is further configured for transmitting, in response to starting a training iteration of the requested FL training, for which a part of the affiliation nodes are not used, affiliation update information including the received identifier of the affiliation entry and information about the affiliation nodes not used for the training iteration. According to another example aspect, the transmitted affiliation information further comprises a predicted FL task completion time. According to another example aspect, the transmitted affiliation information further comprises an event exposure configuration. According to another example aspect, the transmitted affiliation information further comprises FL process information about an FL training. According to another example aspect, the transmitted affiliation information further comprises an event exposure configuration. According to another example aspect, the means for transmitting (260) is further configured for transmitting affiliation retrieve information including FL process information about a planned FL training, the means for receiving (250) is further configured for receiving an affiliation retrieve response comprising an identifier of an FL aggregator corresponding to an on-going FL process that fits the transmitted FL process information. According to a fourth aspect, there is provided a method, comprising transmitting federated learning, FL, task information including characteristics of a requested FL training, receiving a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training, selecting a subset of the set of distributed nodes, indicated by the distributed nodes list, to be used for the requested FL training, and transmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes. According to another example aspect, the method further comprises receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, transmitting, in response to the requested FL training being executed and completed, affiliation delete information including the received identifier of the affiliation entry. According to another example aspect, the method further comprises receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, transmitting, in response to starting a training iteration of the requested FL training, for which a part of the affiliation nodes are not used, affiliation update information including the received identifier of the affiliation entry and information about the affiliation nodes not used for the training iteration. According to another example aspect, the affiliation information further comprises a predicted FL task completion time. According to another example aspect, the affiliation information further comprises an event exposure configuration. According to another example aspect, the affiliation information further comprises FL process information about an FL training. According to another example aspect, the method further comprises transmitting affiliation retrieve information including FL process information about a planned FL training, receiving an affiliation retrieve response comprising an identifier of an FL aggregator corresponding to an on-going FL process that fits the transmitted FL process information. According to another example aspect, transmitting and receiving may be performed via a T1 interface for all methods and apparatuses. According to a fifth aspect, there is provided an apparatus 100 comprising at least one processor 110, at least one memory 120 and at least one interface 130 configured for communication with at least another apparatus, wherein the at least one processor, with the at least one memory and the computer program code and with the at least one interface is configured to cause the apparatus to perform receiving federated learning, FL, task information including characteristics of a requested FL training, to perform generating, in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, distributed nodes for contributing to the requested FL training, and to perform transmitting the distributed nodes list, wherein generating the distributed nodes list is based on a comparison of the characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability Information corresponding to the plurality of distributed nodes, and the processor is further configured to perform in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform receiving affiliation information comprising a list of affiliation nodes, indicating a subset of the set of distributed nodes indicated in the distributed nodes list (S6), creating an affiliation entry including an identifier of the affiliation entry and the list of affiliation nodes, storing the affiliation entry including the identifier of the affiliation entry and the list of affiliation nodes, receiving second FL task information including characteristics of a second requested FL training, generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and transmitting the second distributed nodes list, wherein generating the second distributed nodes list is further based on the stored affiliation entry. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform transmitting the identifier of the affiliation entry, receiving affiliation update information including an identifier of a corresponding affiliation entry, and in response to receiving the affiliation update information, updating the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation update information. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface Is further configured to cause the apparatus to perform transmitting the Identifier of the affiliation entry, receiving affiliation delete information including an identifier of a corresponding affiliation entry, and in response to receiving the affiliation delete information, deleting the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation delete information. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list in response to receiving the affiliation information. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform updating (S17) the stored availability information corresponding to the affiliation nodes in response to receiving the affiliation delete information and / or in response to receiving the affiliation update information. According to another example aspect, the stored availability information corresponding to the plurality of distributed nodes comprises at least one of a status of the plurality of distributed node, or one or more affiliation entries, each indicating a set of distributed nodes selected for a running FL training process. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform storing a list indicating reserved resources of distributed nodes in response to generating the distributed nodes list; receiving second FL task information including characteristics of a second requested FL training, generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes, and transmitting the second distributed nodes list, wherein resources of distributed nodes indicated in the list indicating reserved resources of distributed nodes are excluded from the second distributed nodes list. According to another example aspect, the affiliation information further comprises a predicted FL task completion time, storing the affiliation entry further comprises storing the predicted FL task completion time, and the second distributed nodes list (S35) indicates a set of currently available distributed nodes and further indicates a set of currently unavailable distributed nodes together with an indication of a future time, when the currently unavailable distributed nodes will be available. According to another example aspect, the affiliation information further comprises a first event exposure configuration, the second FL task information further comprises a second event exposure configuration, and the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform evaluating the first event exposure configuration and the second event exposure configuration, and in case the first event exposure configuration and the second event exposure configuration match, transmitting an event report and generating the second distributed nodes list only after the FL training process corresponding to the first event exposure configuration is finished. According to another example aspect, the affiliation information further comprises FL process information about an FL training, storing the affiliation entry further comprises storing the FL process information and an identifier of a corresponding FL aggregator, and the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform receiving affiliation retrieve information or second FL task information, the received information including second FL process information, searching for an affiliation entry with FL process information that fits the second FL process information of the received information, and transmitting an affiliation retrieve response comprising the identifier of a corresponding FL aggregator included in the affiliation entry with FL process information that fits the second FL process information of the received information. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform updating the stored availability information based on a timer per distributed node discovery request and / or periodically. According to a sixth aspect, there is provided an apparatus 200 comprising at least one processor 210, at least one memory 220 and at least one interface 230 configured for communication with at least another apparatus. The processor (e.g., the at least one processor 210, with the at least one memory 220 and the computer program code and with the at least one interface 230) is configured to perform transmitting federated learning, FL, task information including characteristics of a requested FL training, to perform receiving a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training, to perform selecting a subset of the set of distributed nodes, indicated by the distributed nodes list, to be used for the requested FL training, and to perform transmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, transmitting, in response to the requested FL training being executed and completed, affiliation delete information including the received identifier of the affiliation entry. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, transmitting, in response to starting a training iteration of the requested FL training, for which a part of the affiliation nodes are not used, affiliation update information including the received identifier of the affiliation entry and information about the affiliation nodes not used for the training iteration. According to another example aspect, the affiliation information further comprises a predicted FL task completion time. According to another example aspect, the affiliation information further comprises an event exposure configuration. According to another example aspect, the affiliation information further comprises FL process information about an FL training. According to another example aspect, the at least one processor, with the at least one memory and the computer program code and with the at least one interface is further configured to cause the apparatus to perform transmitting affiliation retrieve information including FL process information about a planned FL training, receiving an affiliation retrieve response comprising an identifier of an FL aggregator corresponding to an on-going FL process that fits the transmitted FL process information. According to another example aspect, transmitting and receiving may be performed via a T1 interface for all methods and apparatuses. According to a seventh aspect, there is provided a computer program comprising instructions, which when executed by an apparatus, cause the apparatus to perform a method of this disclosure. According to an eighth aspect, there is provided a non-transitory computer-readable medium comprising program instructions that, when executed by an apparatus, cause the apparatus to perform a method of this disclosure. According to a ninth aspect, there is provided a system, comprising the apparatus according to the first or fifth aspect, and the apparatus according to third or sixth aspect. In the above, various aspects have been described. It should be appreciated that further aspects may be provided by the combination of any two or more of the various aspects described above. Various other aspects are also described in the following detailed description and in the claims. According to some aspects, there is provided subject matter of the independent claims. Some further aspects are defined in the dependent claims. Example embodiments that do not fall under the scope of the claims, if any, are to be interpreted as examples useful for understanding this disclosure. According to some example embodiments, any one or more of the following advantages (e.g., technological improvements, such as technological improvements to computer technology) may be achieved: - Avoiding overload of traditional RAN interfaces via asynchronous signaling for training and / or node selection; Enabling reduced number of signaling for accessing FL capabilities by one or more FL aggregators; Enabling coordination among multiple FL aggregators; - Reducing and removing redundancy in communication for node selection; - Reducing and removing redundancy in training processes; - Avoiding multiple FL processes with similar characteristics to be initiated or trained redundantly; Enabling FL information request negotiations and negotiations for allocating distributed nodes in multiple FL aggregator environments; - Sharing and reusing trained models; Providing parameter updates for further iteration; - Selecting nodes for FL training without affecting existing connectivity; - Allowing to consider a larger set of distributed nodes for a respective node selection process; - Mitigating potential conflicts among multiple FL aggregators initiating node selection for FL training; and / or - Coordinating similar FL training processes. Further details, features, and advantages (e.g., technological improvements, such as technological improvements to computer technology) will be apparent to those skilled in the relevant art(s) from this disclosure. BRIEF DESCRIPTION OF THE DRAWINGS Some example embodiments will be described in greater detail, by way of nonlimiting and illustrative examples, with reference to the drawings (Figs.), in which: Fig. 1 illustrates a process for selecting distributed nodes according to at least some example embodiments of this disclosure; Fig. 2 illustrates a process for selecting distributed nodes according to at least some example embodiments of this disclosure with some added processing steps; Fig. 3 illustrates an optimization of a process for selecting distributed nodes according to at least some example embodiments of this disclosure; Fig. 4 illustrates an optimization of a process for selecting distributed nodes according to at least some example embodiments of this disclosure; Fig. 5 illustrates an optimization of a process for selecting distributed nodes according to at least some example embodiments of this disclosure; Fig. 6 (a) illustrates an example for implementing a database according to at least some example embodiments of this disclosure; Fig. 6 (b) illustrates an alternative example for implementing a database according to at least some example embodiments of this disclosure; and Fig. 7 (a) illustrates an example for implementing an FL aggregator according to at least some example embodiments of this disclosure; and Fig. 7 (b) illustrates an alternative example for implementing an FL aggregator according to at least some example embodiments of this disclosure. DETAILED DESCRIPTION As provided below, the various example embodiments are further described with reference to the accompanying drawings, wherein at least some aspects of the various example embodiments can be freely combined with at least some aspects of other example embodiments of this disclosure, unless described otherwise. However, it is to be understood that the description of the various example embodiments of this disclosure is merely provided by way of non-limiting and illustrative example only, and are not intended to be limiting to any example embodiments of this disclosure or aspects thereof. Moreover, it is to be understood that an apparatus is configured to perform a corresponding method, and a respective method can be executed by a correspondingly configured apparatus although in some cases, only the apparatus or only the method may be described. Federated learning (FL), sometimes also referred to as collaborative learning, can be implemented as a decentralized method for training machine learning models. Federated learning can operate without the need for transmitting data from client devices to a centralized node, such as central server. Federated learning involves deploying AI / ML models directly to the data source instead of transporting the data to the model. This allows, enables, or otherwise facilitates local nodes, like edge devices, to collect data and to train their own model copies, thereby eliminating a further step of transmitting source data to a centralized node. In such approach, raw data on edge devices can be utilized to train the model locally, and thus enhance data privacy. However, selecting distributed nodes to perform local training in an FL process may be complicated and might also lead to a large communication overhead. Example embodiments of this disclosure may relate to the selection of distributed nodes for training in federated learning. Some example embodiments of this disclosure can be applied to various communication systems, including e.g. mobile communication systems (e.g., 5G, 6G) including UE, RAN, CN and O-RAN. Various example embodiments of this disclosure can be applied via various connection technologies, such as air interface, Peer-to-Peer (P2P) interface, or service-based interface (SBI). In this regard, it is also to be understood that certain entities can act as an FL aggregator, e.g., as a central node in addition to as a distributed node performing a federated learning task (e.g., asynchronously, concurrently, and / or simultaneously), or as only an FL aggregator, or as only a distributed node. Examples for both, FL aggregators and distributed nodes, to perform a training task may include user equipment UE, (Radio) Access Network ((R)AN) nodes, Open-RAN (O-RAN) Near-Realtime RAN Intelligent Controllers (RIC), O-RAN NonRealtime RIC, NetWork Data Analytics Functions (NWDAF) and the like (or any combination thereof). In the following, example embodiments will be described as mere examples to illustrate apparatuses, methods and systems according to the various example embodiments of this disclosure, which are not limited to any specific system configuration and / or model. Entities are characterized by their functionalities to act (e.g., as an FL aggregator and / or as a distributed node) in a federated learning process and can be any apparatus involved in a federated learning process. In this regard, it is also to be understood that example embodiments (or aspects thereof) illustrated in different Figures can be freely combined as appropriate. That is, for example, all or only some example embodiments (or aspects thereof) illustrated in any of Figs. 3 to 5 can be freely combined with example embodiments (or aspects thereof) illustrated in Fig. 1. Initially, a baseline model is stored centrally and shared with client devices, allowing them to train models on local data. In the subsequent stage, updates from locally trained models are securely aggregated at the centralized node, hence also called aggregator or more precisely FL aggregator, enhancing the model's generalizability through diverse data sources. The re-trained central model is then after aggregation shared with client devices for iterative improvement in each cycle, ensuring continuous enhancement without compromising privacy. Federated learning may be based on model and data parallelism. Thus, federated learning allows the models to be trained across entities or data sources while preserving privacy and reducing centralized data aggregation and hence, reducing data transmission overhead. FL training process can be explained in the following main steps: • Initialization: A machine learning model (e.g., linear regression, neural network, and so forth) is chosen to be trained on distributed nodes locally and initialized. • Distributed node selection: a fraction of distributed nodes is selected by the FL aggregator to start local training at distributed nodes on local datasets. The selected nodes acquire the current statistical model while the others wait for the next federated round. • Reporting and Aggregation: each selected distributed node transmits its locally trained model information (e.g., delta to initially received model) to the FL aggregator. The FL aggregator aggregates the received models and transmits back model updates and / or an updated model to the nodes. • Termination: once a pre-defined termination criterion is met (e.g., a maximum number of iterations is reached) the FL aggregator aggregates the updates and finalizes a global model that can be redistributed to the distributed nodes. FL integration into the network is an on-going effort in standardization. In the Core Network (CN), FL among multiple NetWork Data Analytics Function (NWDAF) nodes is introduced in SA2 Rei. 18. To achieve optimized operation across the whole network given the privacy and resource (e.g., latency, payload, etc.) requirements, FL can be a candidate machine learning architecture fit for RAN domain as well, especially when considering the distributed and / or disaggregated nature of RAN. For instance, a description of how federated learning can be enabled in RAN is provided. Specifically, an FL aggregator communicates with the distributed nodes to receive their FL capabilities. The FL capabilities are used by the FL aggregator to select a subset of the distributed nodes. FL capabilities may include, for example, local data statistics and characteristics, distributed node characteristics, and a communication channel. However, in such scenario, the FL aggregator communicates with the distributed nodes separately asking for relevant reporting as to whether the distributed nodes can be part of, e.g., contribute to, the training used for selecting a subset of the distributed nodes for FL training process. Such separate communication with the distributed nodes can lead to various drawbacks. Separately asking each distributed node (e.g., a number M =6) about their FL-related capabilities and then selecting a few (number N where M> = N, e.g., N=3 out of M=6) for transmitting FL training parameters could cause resource issues on the respective interfaces impacting RAN / UE-related operations (such as Xn, El, Fl, NG, etc.). For example, an FL aggregator can be a central unit (CU) and distributed nodes can be distributed units (DU). Considering RAN deployments, the number of DUs can be xlOs higher than that of CUs, requiring non-negligible communication resources to select distributed nodes. This can lead to overload of traditional RAN interfaces via asynchronous signaling for training and / or node selection. Furthermore, there can be multiple processes that would like to initiate FL training, e.g., multiple FL aggregators (e.g., FL aggregator 1 and FL aggregator 2 in Fig. 1), in which case such node selection method can grow in size relative to size of the number of processes. In a multiple FL aggregator environment, FL aggregators may compete for the same resources. From a first FL aggregator's node selection process time until the start time of an FL training process, a second FL aggregator can request relevant reporting, from the distributed nodes, a subset of which may overlap with the selected nodes of the first FL aggregator (e.g., selected nodes in Fig. 1). So, when the second FL aggregator selects a set of distributed nodes, the selected set of distributed nodes will be based on an outdated reporting information, causing synchronization and coordination issues. In such case, certain requirements for the FL process may not be met and further iterations for the node selection process for the second FL aggregator may be required. For example, multiple CU-user planes (CU-llps) can be FL aggregators (e.g. an FL aggregator 1 and an FL aggregator 2) and distributed nodes can be DUs (distributed nodes 1 to N) considering the CU-UP and DU cardinality. FL process initiation may be based on slice specific optimization use cases. <Example embodiments> An example process for selection of distributed nodes for federated learning is described with respect to Fig. 1. In at least some example embodiments, there may be provided a shared entity, like a database (DB). For example, the database may be shared across multiple RAN nodes and logical entities, such that the database is accessible by one or more monolithic RAN Nodes as well as the logical entities in disaggregated RAN architecture. In at least some example embodiments, the shared database may communicate with distributed nodes and / or one or more FL aggregators via an interface specific fortraining purposes, such as a T1 interface. Such a separate interface may avoid impact on on-going connectivity and operation related processes (such as Xn, El, Fl, NG, etc.). In at least some example embodiments, the interface can operate over T1AP / SCTP or HTTPS / TCP or HTTP3 / QUIC, or any other suitable protocol stack. In at least some example embodiments, the database may act as a shared entity and may store information about distributed nodes that are generally available to perform training in a federated learning process. That is, the database may store information about distributed nodes with FL training capability. Such distributed nodes with FL training capability are also referred to as a "local trainer". In at least some example embodiments, the database can be a standalone entity and / or a part of an entity. That is, for example in a scenario involving multiple FL aggregators, the database can be incorporated into an FL aggregator while it may as well be a separate node. In a configuration as illustrated in Fig. 1, a single FL aggregator and multiple distributed nodes with FL training capability are considered. In at least some example embodiments, the database may have a storage unit, in which information like characteristics of a plurality of distributed nodes and availability information corresponding to the plurality of distributed nodes is stored. In Fig. 1, an example is considered where this information is already stored. In at least some example embodiments, respective characteristics of the plurality of distributed nodes and availability information corresponding to the plurality of distributed nodes may be transmitted from each distributed node to the database within a registration procedure as shown in Fig. 2 and described in more detail with respect to Fig. 2. Furthermore, according to at least some example embodiments, updated availability information may also be transmitted from distributed nodes to the database (e.g. regularly or in response to certain changes). FL information may, for example, include certain information provided in reporting (e.g., interim or previous training results, availability of computational resources or available training data) as well as an FL type whether an entity may act as an FL aggregator and / or as a distributed node, an FL status and optionally an FL affiliation information as availability information. An FL type (as a part of characteristics of a distributed node) indicates whether the node may act as an "FL aggregator" or a "local trainer" or as both an "FL aggregator" and a "local trainer". Other examples for characteristics of distributed nodes include local data statistics and characteristics, and information about one or more available communication channels. An FL status indicates the current status of a node as to whether the node is currently involved in an FL process (e.g., in an on-going FL training process or in a node selection process for a planned FL training process). For example, the FL status can take values that correspond to "free", "in selection process" and "in training process". "Free" indicates that the node is currently not involved in an FL process. "In selection process" indicates that the node is currently in the node selection part of an FL process (e.g., is included in a distributed nodes list). "In training process" indicates that the node is currently performing local training for an FL training process. FL affiliation information may indicate the FL training operations that a respective node is currently part of. In at least some example embodiments, FL affiliation information may be a list of one or more affiliations, each affiliation of the list of one or more affiliations indicating an on-going FL training process (affiliation information corresponds to a node's on-going FL training processes). Characteristics of a respective distributed node may also include, in addition to the FL type, information about computational resources, such as storage capacity and whether a local trainer comprises or has access to one or more CPUs and / or one or more GPUs. Another example for characteristics of distributed nodes for acting as local trainers is access to data sources or availability and / or generation of training data at the distributed node. In at least some example embodiments, the characteristics of the plurality of distributed nodes and the availability information may be considered to be previously stored instead of performing a registration procedure. When an FL aggregator intends to start an FL training process, according to at least some example embodiments, the FL aggregator transmits, as illustrated in step SI of Fig. 1, e.g., a node discovery request comprising FL task information including characteristics of a requested FL training (e.g., of the intended FL training process). The characteristics of the requested FL training can include FL Information-related requirements like, for example, computational resources, availability time period, area definition, requirements on training data, or the like. In response to receiving the FL task information including characteristics of the requested FL training, a distributed nodes list indicating, out of the plurality of distributed nodes, the database generates a set of distributed nodes for contributing to the requested FL training is generated at the database in step S2. For generating the distributed nodes list, characteristics of the requested FL training and characteristics of the distributed nodes and the availability information of the distributed nodes are assessed. For example, the database may compare computational resources included in the characteristics of the requested FL training with stored information about computational resources of the distributed nodes, e.g., possible local trainers. Accordingly, only distributed nodes with sufficient computational resources to perform a respective local training task are added to the distributed nodes list. Furthermore, generating the distributed nodes list may also be based on stored availability information corresponding to the plurality of distributed nodes, in that only local trainers, whose status is "free", are - by way of example - added to the distributed nodes list. In at least some example embodiments, the database may update the FL status of the distributed nodes added to the distributed nodes list as possible local trainers. For example, in response to adding a local trainer to the distributed nodes list, the database may change the status of the added local trainer from "free" to "in selection process" in step S3. In addition, the database may also set the FL status of the FL aggregator to "in selection process" in response to receiving the FL task information in step SI. In step S4 of Fig. 1, the database transmits the distributed nodes list to the FL aggregator as a response to the received FL task information, e.g., as a response to receiving the node discovery request. In at least some example embodiments, updating the availability information in step S3 may happen after or in response to transmitting the distributed nodes list in step S4. That is, the order of steps S3 and S4 may be swapped. When updating, for example, the FL status as a part of availability information of distributed nodes indicated in the distributed nodes list in step S3, the database may alternatively or additionally assign a status as "reserved resources" to resources of the distributed nodes indicated in the distributed nodes list. The assigned status "reserved resources" can also be considered as availability information. By receiving the distributed nodes list, the FL aggregator receives a list indicating a plurality of distributed nodes, which may be employed as local trainers for the intended and requested FL training process, without previous communication between the FL aggregator and the listed distributed nodes being required. That is, the FL aggregator only needs to query one entity, the database, instead of querying a plurality of distributed nodes. The distributed nodes list may indicate more possible local trainers than required for the requested FL training process. Accordingly, the FL aggregator may select a subset of distributed nodes from the distributed nodes list to perform local training in step S5. This selection may be based, for example, on connectivity between the FL aggregator and the distributed nodes for transmitting the initialized model and receiving locally trained models afterwards. This selection can be based on any known and / or new criteria for selecting local trainers, e.g. based on information about the distributed nodes itself or about available training data at the distributed nodes. Typically, selecting local trainers would be partially related to the characteristics of FL training and characteristics of the distributed nodes. After receiving a distributed nodes list, the FL aggregator may, for example, perform some cross-check of the provided distributed nodes list. Selecting can, for example, be based on some local configuration, dynamic policy, an AI / ML model inference, and / or any prior knowledge about distributed nodes (e.g., distributed node resiliency, popularity for FL training selection, number of UEs served by the node at a given time, etc.). After selecting a subset of local trainers from the distributed nodes list, the FL aggregator transmits FL affiliation information including a list of affiliation nodes indicating the selected nodes to perform local training to the database in step S6. That is, the list of affiliation nodes indicates a subset of the set of distributed nodes indicated in the distributed nodes list. In at least some example embodiments, FL training at the distributed node (e.g., at the local trainers) may be triggered by the FL aggregator according to a procedure in step S7. Alternatively, the database may trigger a start of FL training at the distributed nodes by forwarding FL task information and / or affiliation information. In such scenario, the FL aggregator may transmit an initial model to be trained (e.g., included in the FL task information or included in affiliation information). In Fig. 2, a process for selecting distributed nodes according to at least some example embodiments of this disclosure with some added processing steps is illustrated. The process illustrated in Fig. 2 comprises all steps illustrated in Fig. 1 and some additional steps like processing and transmission of information. All steps of Fig. 2, which are also illustrated in Fig. 1, are the same as described with respect to Fig. 1 and a repeated description of steps illustrated in Fig. 1 is thus omitted. According to at least some example embodiments, the FL aggregator and the plurality of distributed nodes may be registered at the database in a first step Sil. An FL aggregator and distributed nodes to be registered transmit their respective FL information (e.g., within an FL capability create request) to the database. The FL information may, for example, include certain information provided in reporting (e.g., interim or previous training results, availability of computational resources or available training data) as well as an FL type indicating whether an entity may act as an FL aggregator and / or as a distributed node, an FL status and optionally an FL affiliation information as some availability information. In response to receiving an FL capability request, the database creates corresponding entries for registered entities and responds back with an FL Create Response to notify the registered distributed node or FL aggregator about successful registration or unsuccessful registration operation with, for example, a proper failure cause value. An FL type (as a part of characteristics of a distributed node) indicates whether the node may act as an "FL aggregator" or a "local trainer" or as both an "FL aggregator" and a "local trainer". Other examples for characteristics of distributed nodes include local data statistics and characteristics, and information about one or more available communication channels. An FL status indicates the current status of a node as to whether the node is currently involved in an FL process (e.g., in an on-going FL training process or in a node selection process for a planned FL training process). For example, the FL status can take values that correspond to "free", "in selection process" and "in training process". "Free" indicates that the node is currently not involved in an FL process. "In selection process" indicates that the node is currently in the node selection part of an FL process. "In training process" indicates that the node is currently performing local training for an FL training process. FL affiliation information may indicate the FL training operations that a respective node is currently part of. In at least some example embodiments, FL affiliation information may be a list of one or more affiliations, each affiliation indicating an on-going FL training process (e.g., affiliation information may correspond to a node's on-going FL training processes). A capability create request transmitted in step Sil may comprise further characteristics of a respective distributed node (e.g., information about computational resources such as storage capacity and whether a local trainer comprises or has access to one or more CPUs and / or one or more GPUs). Another example of characteristics of distributed nodes acting as local trainers is access to data sources or availability and / or generation capabilities of training data at the distributed node (e.g., at the local trainer). In the example illustrated in Fig. 2, selection of distributed nodes to perform local training may be implemented as described above with respect to Fig. 1. After selecting a subset from the set of distributed nodes (e.g., selecting a subset of local trainers, from the distributed nodes list), the FL aggregator transmits FL affiliation information including a list of affiliation nodes indicating the selected nodes to perform local training to the database in step S6 as described with respect to step S6 of Fig. 1. The FL affiliation information indicates the local trainers selected from the distributed nodes list and computational resources requested at the indicated local trainers. Additionally, the affiliation information may also comprise FL training process related information (e.g., a training description, etc.). In addition, in response to receiving the list of affiliation nodes indicating the selected nodes to perform local training, the database creates an affiliation entry in step S12 according to at least some example embodiments. The affiliation entry according to at least some example embodiments of this disclosure comprises at least an affiliation ID and a list of distributed nodes involved in the FL process. The database assigns an affiliation identifier to identify the FL training process that is to be started at the affiliation nodes. The affiliation entry includes information that may have been included in the FL task information and / or the FL affiliation information, such as an FL aggregator identifier or identifiers of distributed nodes selected to perform local training. If provided by the FL aggregator in the FL task information and / or the FL affiliation information, the affiliation entry may also include FL training process related information, such as a training description, initial model architecture or the like. As is illustrated in Fig. 2, the database may further, in response to receiving a list of affiliation nodes indicating the selected nodes to perform local training, update the FL status of the distributed nodes indicated in the distributed nodes list in step S13. In particular, for distributed nodes indicated in the distributed nodes list but not indicated in the list of affiliation nodes, the database may change the FL status from "in selection process" to "free". If a distributed node indicated In the distributed nodes list but is not indicated in the list of affiliation nodes from a selection process by another FL aggregator, its status may be maintained to be "in selection". For distributed nodes indicated in the affiliation information, the database may change the status, for example, to "in training process". Additionally, the database may set the status of the FL aggregator to "in FL process" in response to receiving the affiliation information from the FL aggregator. If a status like "reserved resources" was assigned to resources of distributed nodes indicated in the distributed nodes list when updating the availability information in step S3, the status like "reserved resources" associated with the FL aggregator is removed for distributed nodes not indicated in the list of affiliation nodes while the status "reserved resources" associated with the FL aggregator is also removed for distributed nodes or resources of distributed nodes indicated in the affiliation information as it is changed for example to a status like "in training process". Furthermore, the FL status may also comprise information about available resources of registered distributed nodes. In such case, the FL aggregator may also indicate specific recourses of distributed nodes indicated in the list of affiliation nodes. In at least some example embodiments, indicated resources of selected distributed nodes indicated in the list of affiliation nodes may be assigned a specific status (e.g., "requested computational resources"). Resources stored as available resources of non-selected distributed nodes are left as available resources. Upon creating an affiliation entry in step S12, the database transmits an affiliation create response to the FL aggregator in step S14. The affiliation create response comprises the affiliation identifier to identify (e.g., uniquely identify) the FL training process. After the FL training process is finished (step S15), the FL aggregator transmits an FL affiliation delete request (e.g., affiliation delete information) to the database to notify the database about the finished FL training process in step S16. The affiliation delete request comprises the affiliation identifier to identify the FL training process. In response to receiving the affiliation delete request, the database also updates the FL status of all distributed nodes of the list of distributed nodes involved in the FL process, included in the affiliation entry, to "free" (S17), if these distributed nodes are not in an FL training process by another FL aggregator. Furthermore, in at least some example embodiments including a status (e.g., "requested computational resources"), assigned to resources of selected distributed nodes indicated in the list of affiliation nodes, this status is, for resources of distributed nodes and / or distributed nodes indicted in the affiliation entry, changed to "available resources" in response to receiving the affiliation delete request. In response to receiving the affiliation delete request (e.g., affiliation delete information), the database also deletes the affiliation entry, to which the affiliation identifier included in the affiliation delete request is assigned (S18). Upon deleting the affiliation entry, the database may transmit an FL Affiliation delete response to the FL aggregator (S19). During the FL training, there can be multiple iterations (e.g., distributed nodes train locally, FL aggregator aggregates the local models and decides to do another iteration of FL training). In subsequent iterations, FL aggregator may choose to not include (e.g., omit, refrain from including, etc.) some of the previously selected distributed nodes. In this case, alternatively or in addition to sending affiliation delete information in step S16, the FL aggregator may transmit an FL affiliation update request (e.g., affiliation update information) to the database. Such FL affiliation update request (e.g. affiliation update information) may indicate a modification an affiliation information transmitted in step S6, (e.g., if a part of the selected affiliation nodes were used fortraining in a first iteration but are not used fortraining in a subsequent or second iteration). For example, the affiliation update information may comprise - according to at least some example embodiments, a corresponding affiliation identifier and a list of previous affiliation nodes that are not used for subsequent iterations. In such a scenario, the database would then update the affiliation entry with the received modification information, that is, update the affiliation entry and / or the stored availability information of distributed nodes that were affiliation nodes in previous iterations but are not used for FL training any more in the subsequent iterations. For example, the database may remove distributed nodes indicated in the list of previous affiliation nodes that are not used for subsequent iterations included in the affiliation update information from the affiliation entry with the corresponding affiliation identifier. Upon updating the affiliation entry, the database may transmit an FL Affiliation update response to the FL aggregator. In at least some example embodiments, FL status information may be updated for example once after node selection process, once after start of an FL training process, once after termination of an FL training process and in case of temporarily stopping the FL training. In at least some example embodiments, updating the FL status can be done in response to information transmitted from a node itself to the database or in certain cases, updating the FL status may be triggered by the database in response to receiving information, such as FL affiliation information and / or an FL affiliation delete request. In at least some example embodiments, there may be a scenario considered, involving multiple FL aggregators. For example in Fig. 3, a scenario involving two FL aggregators (e.g., FL aggregator 1 and FL aggregator 2) is illustrated. In particular, in a scenario illustrated in Fig. 3, a first FL aggregator (e.g., FL aggregator 1) discovers some distributed nodes via interaction with the database. That is, the database generates and transmits the distributed nodes list as described with respect to steps SI to S4 in one of Figs. 1 or 2 to the FL aggregator 1. However, in Fig. 3, a scenario is illustrated including a point in time when the database has not yet created the affiliation entry (e.g., the FL aggregator 1 has not yet sent affiliation information in step S6 and also the FL training is not yet started). At the point in time described above, it is in Fig. 3 considered that a second FL aggregator (e.g., FL aggregator 2) would like to discover some distributed nodes. That is, after the FL aggregator 1 received the distributed nodes list for his requested FL training (e.g., first FL training), but before the requested first FL training has started, the FL aggregator 2 transmits second node discovery request comprising second FL task information to the database in step S21. In such scenario, there may be the case that the database can find some distributed nodes that are common possible local trainers for the two FL aggregators. However, only looking at the available resources can cause misinformation regarding the distributed node's actual resources if FL aggregator 1 starts the FL affiliation creation later than the FL aggregator 2 transmits the second FL task information. In the scenario illustrated in Fig. 3, steps SI', S2' and S4' are the same as steps SI, S2 and S4, respectively, described with respect to Fig. 1. That is, in step SI, FL aggregator 1 transmits first FL task information corresponding to a requested first FL training. In step S2', the database generates a first distributed nodes list similar to generating the distributed nodes list in step S2 of Fig. 1. According to at least some example embodiments, the database may when updating the availability information in step S3 also store information about reserved resources. When updating (e.g. the FL status as a part of availability information of distributed nodes indicated in the distributed nodes list in step S3'), the database may also assign a status as "reserved resources" to resources of the distributed nodes Indicated in the distributed nodes list according to at least some example embodiments. Alternatively or in addition, the database may, in step S3', store a list indicating reserved resources of distributed nodes (e.g., information about reserved resources) indicated in the first distributed nodes list. The list indicating reserved resources may comprise an identifier of the corresponding FL aggregator (e.g., FL aggregator 1) and information about resources allocated for the first FL task. In Fig. 3, it is considered that after the database transmits the first distributed nodes list in step 54' to the FL aggregator 1, FL aggregator 2 transmits second FL task information corresponding to a requested second FL training afterwards (step S21). In response to receiving the second FL task information including characteristics of the requested second FL training in step S21, a second distributed nodes list indicating, out of the plurality of distributed nodes, a second set of distributed nodes for contributing to the requested second FL training is generated at the database in step S22. For generating the second distributed nodes list, characteristics of the requested second FL training and characteristics of the distributed nodes and the availability information of the distributed nodes are assessed. For example, the database may compare computational resources included in the characteristics of the requested second FL training with stored information about computational resources of the distributed nodes (e.g., possible local trainers). Accordingly, only distributed nodes with sufficient and / or suitable computational resources to perform a respective local training task are added to the distributed nodes list. Furthermore, generating the distributed nodes list may also be based on the stored availability information corresponding to the plurality of distributed nodes, in that e.g. only distributed nodes, whose status is "free", are added to the distributed nodes list. By considering only distributed nodes, whose FL status (e.g., availability Information) is "free" for generating the second distributed nodes list, conflicts can be mitigated in a multiple FL aggregator environment. Alternatively or in addition, in at least some example embodiments, the database may, when generating the second distributed nodes list in step S22 can exclude from the second distributed nodes list, all distributed nodes or resources of distributed nodes indicated in the list indicating reserved resources of distributed nodes (e.g., distributed nodes or resources of distributed nodes indicated in the first distributed nodes list). Such exclusion of (potentially) non-available resources (e.g., distributed nodes or (partial) resources of distributed nodes), from the second distributed nodes list also allows mitigating possible conflicts in a multiple FL aggregator environment. According to at least some example embodiments, additional optimization in node selection may allow, enabling, or otherwise facilitate FL task negotiations. In particular, it could be the case that fulfilling a requested FL task is not possible (e.g., due to a lack of available resources considering reserved resources in a multiple FL aggregators environment). In such a case, FL task negotiations may be applied in at least some example embodiments. To allow, enable, or otherwise facilitate FL task negotiations, an FL aggregator according to at least some example embodiments can provide alternative requirements for the requested resources during node discovery as well as an approximate FL training process finish time included in the FL affiliation information as transmitted to the database. That is, such alternative requirements can be, for example, included in FL task information. In Fig. 4, a scenario is assumed, in which FL aggregator 1 received a first distributed nodes list similar to step S4 in Figs. 1 and 2 and similar to step 54' in Fig. 3. Afterwards, like the FL aggregator in step S5 of Figs. 1 and 2, the FL aggregator 1 in Fig. 4 selects distributed nodes to perform local training (e.g., affiliation nodes) from the set of distributed nodes indicated in the first distributed nodes list. After selecting distributed nodes to perform local training (e.g., affiliation nodes) from the set of distributed nodes indicated in the distributed nodes list, FL aggregator 1 transmits affiliation information indicating the affiliation nodes to the database (step S31). According to at least some example embodiments, the affiliation information comprises, in addition to a list of affiliation nodes indicating the selected nodes to perform local training to the database, a predicted FL task completion time of the corresponding FL task. In at least some example embodiments, the predicted finish time can also be updated by a corresponding FL aggregator (e.g. FL aggregator 1) depending on an FL cycle (via e.g. an FL affiliation update request). The predicted FL task completion time indicates at what time the reserved resources will become available again. In response to receiving the list of affiliation nodes indicating the selected nodes to perform local training, the database creates an affiliation entry in step S32 according to at least some example embodiments. The affiliation entry according to at least some example embodiments of this disclosure comprises an affiliation ID and a list of distributed nodes involved in the FL process. The database assigns an affiliation identifier to identify the affiliation entry corresponding to the FL training process that is to be started at the affiliation nodes. The affiliation entry can include information that may have been included in the FL task information and / or the FL affiliation information, such as an FL aggregator identifier, or identifiers of distributed nodes selected to perform local training. If provided by the FL aggregator (e.g., FL aggregator 1) in the FL task information and / or the FL affiliation information, the affiliation entry may also include FL training process related information, such as a training description, or the like. In at least some example embodiments, the affiliation entry further comprises the predicted FL task completion time. Upon creating an affiliation entry in step S32, the database may transmit an affiliation create response to the FL aggregator in step S33. The affiliation create response comprises the affiliation identifier to identify the FL training process. Here, it is assumed in Fig. 4 that a second FL aggregator (e.g., FL aggregator 2) transmits second FL task information corresponding to a requested second FL training in step S34. According to at least some example embodiments, this second FL task information includes characteristics of a requested second FL training, a combination of requested currently available resources and requested future available resources of the distributed nodes, as alternative requirements. This could be relevant if the FL aggregator (e.g., FL aggregator 2) would like to consider a larger set of distributed nodes for its selection process (e.g., not based only on their current available resources but also the future available resources). In such a scenario, FL aggregator 2 may adjust its FL training process (e.g., time to complete, etc.) according to the current and future availability of the resources on distributed nodes. For example, FL aggregator 2 can consider to reserve 10% CPU of distributed node 1 until a time T1 and 50% CPU of distributed node 1 after the time T1 until a time T2. Alternatively or in addition, an FL aggregator (e.g., FL aggregator 2) can also set an event exposure configuration to be updated based on the available resources of distributed nodes (e.g. periodically and / or based on a timer, that is, based on a timer per distributed node discovery request), in regular intervals or in a determined time period. The event exposure configuration can, for example, indicate a time period requested or required for respective computational resources. In the event exposure configuration, the indicated time period can be adapted in accordance with availability of computational resources (e.g., the indicated time period can be increased if fewer computational resources are available and the opposite way). In step S35, the database generates the second distributed nodes list, indicating, out of the plurality of distributed nodes, a set of second distributed nodes for contributing to the requested second FL training. For generating the second distributed nodes list, characteristics of the requested second FL training and characteristics of the distributed nodes and the availability information of the distributed nodes are assessed. For example, the database may compare computational resources included in the characteristics of the requested second FL training with stored information about computational resources of currently available distributed nodes, (e.g., distributed nodes with an FL status set to "free"). In at least some example embodiments, generating the second distributed nodes list may further be based on the predicted FL task completion time. That is, the second distributed nodes list may also indicate a set of currently unavailable distributed nodes (e.g., distributed nodes with a status "in training process" together with an indication of a future time), when the currently unavailable distributed nodes will be available. Hence, the database can use the predicted FL task completion time to provide a distributed nodes list based on currently available resources and based on currently reserved (e.g., currently unavailable resources), in handling node discovery request and preparing a list of distributed nodes accordingly. Furthermore, the database may also set or adapt an event exposure configuration. In step S36, the database transmits the second distributed nodes list indicating a set of currently available distributed nodes and further indicating a set of a currently unavailable distributed nodes (e.g., distributed nodes with a status "in training process" together with an indication of a future time), when the currently unavailable distributed nodes will be available. Together with the FL status (e.g., "in selection process", "in training process", "free") information and / or the predicted FL task completion time, the second FL aggregator (e.g., FL aggregator 2) can understand that more resources of the given distributed nodes will be available in the future and can also take this into consideration when selecting the distributed nodes. Furthermore, in case an event exposure configuration is set, the database may detect one or more events provided in the event exposure configuration. That is, if the database detected one or more events, the database transmits an FLTraining Node Discovery Notification message. The FL aggregator (e.g., FL aggregator 2) can process the provided information and take any further action necessary together with communication with the database. According to at least some example embodiments, the event exposure configuration can include event characteristics (e.g., any subset of availability information) and reporting characteristics (such as periodic notification (e.g., every hour), threshold-based notification (e.g., min available resources value, etc.). When the database stores such event configuration, it may set up an event detection mechanism. When an event is observed according to the event exposure configuration, the database transmits a response including an event report for the observed event. Based on such an event detection mechanism involving event exposure configuration, the database does not always initiate a node discovery request by a second FL aggregator but can instead wait until the availability of nodes changes in the direction that the second FL aggregator requested previously. Hence, said event detection mechanism may, for example, be optimized (e.g., reduces the signaling amount and frequency). According to at least some example embodiments, a node selection process for selecting distributed nodes (e.g., local trainers) for an FL training task can be further or otherwise optimized for coordination of FL training processes. For instance, there may be scenarios, in which more than one FL aggregators are interested in FL training with a similar scope or description (e.g., in a similar FL training process). According to at least some example embodiments, the database can provide information on on-going FL training processes to the FL aggregators via FL affiliation entries (e.g., reading FL affiliation entries). Below, it is described an example embodiment, in which affiliation information further includes information about a scope or a description of an FL training process. However, the scope and / or description of the FL training process can also be included in the FL task information transmitted (e.g., in step SI of Fig. 1). In Fig. 5, a scenario is assumed, in which FL aggregator 1 received a first distributed nodes list similar to step S4 in Figs. 1 and 2 and similar to step 54' in Fig. 3. Afterwards, like the FL aggregator in step S5 of Figs. 1 and 2, the FL aggregator 1 of Fig. 5 selects distributed nodes to perform local training (e.g., affiliation nodes) from the set of distributed nodes indicated in the distributed nodes list. After selecting distributed nodes to perform local training (e.g., affiliation nodes) from the set of distributed nodes indicated in the distributed nodes list, the FL aggregator 1 transmits affiliation information to the database (step S41). According to at least some example embodiments, the affiliation information further comprises, FL process information including, for example, a scope of the FL training process and / or a description of the FL training process. The description of the FL training process can be a training scope defined by an identifier or a string text, any other meaningful format, etc. The FL process information can, for example, also indicate an initial model to be trained and / or aims of what a trained model will be capable to output or predict. In response to receiving the list of affiliation nodes indicating the selected nodes to perform local training, the database creates an affiliation entry in step S42. The affiliation entry according to at least some example embodiments of this disclosure comprises an affiliation ID and a list of distributed nodes involved in the FL training process. The database assigns an affiliation identifier to identify the affiliation entry corresponding to the FL training process that is to be started at the affiliation nodes. The affiliation entry can include information that may have been included in the FL task information and / or in the FL affiliation information, such as an FL aggregator identifier, or identifiers of distributed nodes selected to perform local training. If provided by the FL aggregator in the FL task information and / or the FL affiliation information, the affiliation entry can also include FL training process related information, such as a training description, or the like. Upon creating an affiliation entry in step S42, the database transmits an affiliation create response to the FL aggregator in step S43. The affiliation create response comprises the affiliation identifier to identify the affiliation entry corresponding to the FL training process. In Fig. 5, it is assumed that after the affiliation entry for the FL training process requested by FL aggregator 1 is created, a second FL aggregator wants to (e.g., is to, is caused to) start an FL training process. According to at least some example embodiments, FL aggregators may coordinate similar FL training processes. Therefore, before requesting information about available distributed nodes, an FL aggregator (e.g., FL aggregator 2) transmits an FL affiliation retrieve request (e.g., affiliation retrieve information) including second FL process information to the database (step S44). The second FL process information of the FL affiliation retrieve request includes, for example, a scope of the FL training process and / or a description of the FL training process. The description of the FL training process can be a training scope defined by an identifier or a string text, etc. According to at least some example embodiments, affiliation retrieve information Including second FL process information may also be included in second FL task information corresponding to a requested second FL training transmitted by the second FL aggregator (e.g., FL aggregator 2), like in step S34 of Fig. 4. In such a scenario, the steps S45 to S47 described below may be performed equivalently (or similarly) after the second FL aggregator transmits the affiliation retrieve information within the second FL task information instead of within an FL affiliation retrieve request. In response to receiving the FL affiliation retrieve request, the database compares the second FL process information of the FL affiliation retrieve request with FL process information of stored affiliation entries (e.g. FL process information of the affiliation entry created in step S42). By this comparison, the database checks whether there is an on-going FL training process that fits the scope included in the second FL process information of the FL affiliation retrieve request (e.g., the database checks whether there is an on-going FL training process similar to the FL training process FL aggregator wants to start (step S45)). That is, the database searches for an affiliation entry with FL process information that fits the second FL process information of the FL affiliation retrieve request. If the database finds an on-going FL training process that fits the scope included in the second FL process information of the FL affiliation retrieve request in step S45, the database subsequently transmits an FL affiliation retrieve response to the FL aggregator 2 (step S46). The FL affiliation retrieve response includes information about the found on-going FL training process that fits the scope included in the second FL process information of the FL affiliation retrieve request, such as a list of affiliations, e.g., a list of distributed nodes performing local training in the on-going FL training process, and / or FL aggregator information like an identifier (e.g., FL aggregator l's identifier) of the FL aggregator coordinating the on-going FL training process. That is, the database transmits an affiliation retrieve response comprising the Identifier of a corresponding FL aggregator included in the affiliation entry with FL process information that fits the second FL process information of the FL affiliation retrieve request. According to at least some example embodiments, the above-described process for transmitting an affiliation retrieve response is based on an authorization of the FL aggregator 2 to access and / or initiate FL operations of the FL aggregator 1. In at least some example embodiments, the database can implement such authorization mechanism (similar to an NRF in a 5G CN). Alternatively, the database can delegate this to another entity that implements such a mechanism. Afterwards, that is after the FL aggregator 2 receives the FL affiliation retrieve response, the FL aggregators (e.g., FL aggregator 1 and FL aggregator 2) can coordinate on the training process. For example, FL aggregator 1 can share the trained model, whereas FL aggregator 2 can provide model parameter updates for further iteration, or the like. That is, the second FL aggregator (e.g., FL aggregator 2) receives an affiliation retrieve response comprising an identifier of an FL aggregator corresponding to an on-going FL process that fits the transmitted FL process information. Receiving this affiliation retrieve response allows, enables, or otherwise facilitates the FL aggregator 2 to contact the FL aggregator corresponding to an on-going FL process that fits the transmitted FL process information (e.g., FL aggregator 1) in order to subsequently coordinate on the training process. Additionally, in at least some example embodiments, any node can update or even delete its information registered at the database. This can be allowed, enabled or otherwise facilitated by FL Capability Update and FL Capability Delete procedures. In at least some example embodiments, a node can also retrieve a specific distributed node's information via an FL Capability Retrieve procedure. While in case T1 operates over a point-to-point interface (e.g., T1AP / SCTP) this interaction uses request / response communication paradigm, in case T1 operates over SBI (e.g., HTTPS / TCP, HTTP / QUIC) this interaction uses request / response or subscribe / notify communication paradigm. This may also apply to FL affiliation create, FL affiliation update, FL affiliation delete, and FL affiliation retrieve procedures. Any features of the above-described procedures and functions may be implemented by respective functional elements, processors, or the like, as described below. In general terms, the respective devices / apparatuses (and / or parts thereof) may represent means for performing respective operations and / or exhibiting respective functionalities, and / or the respective devices (and / or parts thereof) may have functions for performing respective operations and / or exhibiting respective functionalities. When it is stated that the processor (or some other means) is configured to perform some function, this is to be construed to be equivalent to a description stating that at least one processor, potentially in cooperation with computer program code stored in the memory of the respective apparatus, is configured to cause the apparatus to perform at least the thus mentioned function. Also, such function is to be construed to be equivalently implementable by specifically configured means for performing the respective function (e.g., the expression "processor configured to [cause the apparatus to] perform xxx-ing" is construed to be equivalent to an expression such as "means for xxx-ing"). Alternatively, such function is to be construed to be equivalently implementable by units specifically configured to perform the respective function (e.g., the expression "processor configured to [cause the apparatus to] perform xxx-ing" is construed to be equivalent to an expression such as "xxx unit configured to xxx"). In Fig. 6, two examples, one in each of Figs. 6a and 6b, for implementing a database 100, 100' according to at least some example embodiments of this disclosure are illustrated. According to at least some example embodiments, an apparatus representing the database (DB) comprises at least one processor 110, at least one memory 120 and at least one interface 130 configured for communication with at least another apparatus. The processor (e.g., the at least one processor 110, with the at least one memory 120 and the computer program code and with the at least one interface 130) is configured to perform receiving federated learning, FL, task information including characteristics of a requested FL training (thus the apparatus comprising corresponding means for receiving 150 federated learning, FL, task information including characteristics of a requested FL training), to perform generating, in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, distributed nodes for contributing to the requested FL training (thus the apparatus comprising corresponding means for generating 170, in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, distributed nodes for contributing to the requested FL training), and to perform transmitting the distributed nodes list (thus the apparatus comprising corresponding means for transmitting 160 the distributed nodes list), wherein generating the distributed nodes list is based on a comparison of the characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and the processor (e.g., the at least one processor 110, with the at least one memory 120 and the computer program code and with the at least one interface 130) is further configured to perform in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list (thus the apparatus comprising corresponding storing means 180 for in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list). Alternatively, according to at least some example embodiments of this disclosure, the database can be an apparatus, comprising means for receiving 150 federated learning, FL, task information including characteristics of a requested FL training, and means for generating 170, in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, a set of distributed nodes suitable for contributing to the requested FL training, and means for transmitting 160 the distributed nodes list, wherein generating the distributed nodes list is based on a comparison of the characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and the apparatus further comprises storing means 180 for, in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list. In Fig. 7, two examples, one in each of Figs. 7a and 7b, for implementing an FL aggregator according to at least some example embodiments of this disclosure are illustrated. According to at least some example embodiments, an apparatus representing the FL aggregator 200, 200' comprises at least one processor 210, at least one memory 220 and at least one interface 230 configured for communication with at least another apparatus. The processor (e.g., the at least one processor 210, with the at least one memory 220 and the computer program code and with the at least one interface 230) is configured to perform transmitting federated learning, FL, task information including characteristics of a requested FL training (thus the apparatus comprising corresponding means for transmitting 260 federated learning, FL, task information including characteristics of a requested FL training), to perform receiving a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training (thus the apparatus comprising corresponding means for receiving 250 a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training), to perform selecting a subset of the set of distributed nodes, indicated by the distributed nodes list, to be used for the requested FL training (thus the apparatus comprising corresponding means for selecting 270 a subset of the set of distributed nodes, indicated by the distributed nodes list, to be used for the requested FL training), and to perform transmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes (thus the apparatus comprising corresponding means for transmitting 260, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes). Alternatively, according to at least some example embodiments, the FL aggregator 200' can be implemented as an apparatus, comprising means for transmitting 260 federated learning, FL, task information including characteristics of a requested FL training, means for receiving 250 a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training, means for selecting 270 a subset of the set of distributed nodes, indicated by the received distributed nodes list, to be used for the requested FL training, wherein the means for transmitting 260 is further configured for transmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes. Furthermore, it is to be understood that when it is stated that the processor (or some other means) is configured to perform some function, such function is to be construed to be equivalently implementable by specifically configured circuitry (e.g., the expression "processor configured to [cause the apparatus to] perform xxx-ing" is construed to be equivalent to an expression such as "xxx-circuitry configured to perform xxx-ing"). As used herein, the term "circuitry" may refer to one or more or all of the following: (a) hardware-only circuit implementations (such as implementations in only analog and / or digital circuitry) and (b) combinations of hardware circuits and software, such as (as applicable): (i) a combination of analog and / or digital hardware circuit(s) with software / firmware and (ii) any portions of hardware processor(s) with software (comprising digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions) and (c) hardware circuit(s) and or processor(s), such as a microprocessor(s) or a portion of a microprocessor(s), that requires software (e.g., firmware) for operation, but the software may not be present when it is not utilized for operation. This definition of circuitry applies to all uses of this term herein, comprising in any claims. As a further example, as used herein, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in a server, a cellular network device, or other computing or network device. For further details regarding the operability / functionality of the individual apparatuses, reference is made to the above description in connection with any one of Figures 1 to 7, respectively. For the purpose of this disclosure as described herein above, it should be noted that: - method steps may be implemented as software code portions and run using at least one processor at a network server or network entity (as examples of devices, apparatuses and / or modules thereof, or as examples of entities comprising apparatuses and / or modules therefore), and may further be implemented as software code independent and can be specified using any known or future developed programming language as long as the functionality defined by the method steps is preserved; - generally, any method step may be implemented as software and / or by hardware without changing the example embodiments and its modification in terms of the functionality implemented; - method steps and / or devices, units or means likely to be implemented as hardware components at the above-defined apparatuses, or any module(s) thereof, (e.g., devices carrying out the functions of the apparatuses according to the example embodiments as described herein) are hardware independent and can be implemented using any known or future developed hardware technology or any hybrids of these, such as MOS (Metal Oxide Semiconductor), CMOS (Complementary MOS), BiMOS (Bipolar MOS), BiCMOS (Bipolar CMOS), ECL (Emitter Coupled Logic), TTL (Transistor-Transistor Logic), etc., using for example ASIC (Application Specific IC (Integrated Circuit)) components, FPGA (Field-programmable Gate Arrays) components, CPLD (Complex Programmable Logic Device) components or DSP (Digital Signal Processor) components; - an apparatus like the database or a distributed node, local trainer, etc. may be implemented by a semiconductor chip, a chipset, or a (hardware) module comprising such chip or chipset; this, however, does not exclude the possibility that a functionality of an apparatus or module, instead of being hardware implemented, be implemented as software in a (software) module such as a computer program or a computer program product comprising executable software code portions for execution / being run on a processor; - an entity may be regarded as an apparatus or as an assembly of more than one apparatus, whether functionally in cooperation with each other or functionally independently of each other but in a same device housing, for example. In general, it is to be noted that respective functional blocks or elements according to above-described aspects can be implemented by any known means, either in hardware and / or software, respectively, if it is only adapted to perform the described functions of the respective parts. The mentioned method steps can be realized in individual functional blocks or by individual devices, or one or more of the method steps can be realized in a single functional block or by a single device. Generally, any method step is suitable to be implemented as software or by hardware without changing this disclosure. Devices and means can be Implemented as individual devices, but this does not exclude that they also can be implemented in a distributed fashion throughout the system, as long as the functionality of the device is preserved. Software in the sense of this description comprises software code as such comprising code means or portions ora computer program or a computer program product for performing the respective functions, as well as software (or a computer program or a computer program product) embodied on a tangible medium, such as a computer-readable (storage) medium having stored thereon a respective data structure or code means / portions or embodied in a signal or in a chip, potentially during processing thereof. This disclosure also covers any conceivable combination of method steps and operations described above, and any conceivable combination of nodes, apparatuses, modules or elements described above, as long as the abovedescribed concepts of methodology and structural arrangement are applicable. Even though the disclosure is describes various example embodiments with reference to the accompanying drawings, it is to be understood that the disclosure is not restricted thereto. Rather, it is apparent to those skilled in the relevant art(s) that this disclosure can be modified in various ways without departing from the scope of the various example embodiments disclosed herein. According to at least some example embodiments, there may be provided an apparatus comprising at least one processor and at least one memory (e.g., non-transitory computer readable medium) storing instructions that, when executed by the at least one processor, cause the apparatus to perform at least any method of this disclosure. The term "non-transitory," as used herein, is a limitation of the medium itself (e.g., tangible, not a signal) as opposed to a limitation on data storage persistency (e.g., RAM vs. ROM). According to at least some example embodiments, there may be provided a computer program comprising instructions which, when executed by an apparatus, cause the apparatus to perform at least any method of this disclosure. As used herein, the terms "first X" and "second X" include the options that "first X" is the same as "second X" and that "first X" is different from "second X", unless otherwise specified. These terms are merely used to distinguish one element from another without indicating a temporal relationship, unless otherwise apparent from the disclosure. As used herein, "at least one of the following: " and "at least one of " and similar wording, where the list of two or more elements are joined by "and" or "or," mean at least any one of the elements, or at least any two or more of the elements, or at least all the elements. As used herein, the expression "and / or" also includes any and all combinations of the listed terms, including at least any one of the elements, or at least any two or more of the elements, or at least all of the elements. As used herein, the term "or" refers to a non-exclusive "or" unless otherwise indicated (e.g., use of "or else" or "or in the alternative"). As used herein, unless stated explicitly, performing a respective feature, step, or functionality "in response to A" does not indicate that the respective feature, step, or functionality is performed immediately after "A" occurs as one or more unstated and intervening features, steps, or functionalities may be performed (at least in part) between an occurrence of the respective feature, step, or function and "A". Analogously, performing a respective feature, step, or functionality "based on A" does not indicate that the respective feature, step, or functionality is performed solely based on "A" as the respective feature, step, or functionality may be further based on one or more unstated features, steps, or functionalities in addition to "A". PARTIAL LIST OF ABBREVIATIONS AI - Artificial Intelligence CN - Core Network CP - Control Plane CU - Central Unit DB - Database DU - Distributed Unit FL - Federated Learning HTTP - Hypertext Transfer Protocol 5 HTTPS - Hypertext Transfer Protocol Secure ML - Machine Learning NWDAF - Network Data Analytics Function P2P - Peer-to-Peer QUIC - Quick UDP Internet Connections 10 RAN - Radio Access Network SBA - service-based architecture SBI - service-based interface SCTP - Stream Control Transmission Protocol TCP - Transmission Control Protocol 15 UE - User Equipment UP - User Plane

Claims

1. An apparatus (100'), comprising:means for receiving (150) federated learning, FL, task information including characteristics of a requested FL training, andmeans for generating (170), in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, a set of distributed nodes for contributing to the requested FL training, and means for transmitting (160) the distributed nodes list, wherein generating the distributed nodes list is based on a comparison of characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and based on stored availability information corresponding to the plurality of distributed nodes, and the apparatus further comprisesstoring means (180) configured for, in response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list.

2. The apparatus according to claim 1, whereinthe means for receiving (150) is further configured for receiving affiliation information comprising a list of affiliation nodes, indicating a subset of the set of distributed nodes indicated in the distributed nodes list,the storing means (180) is further configured for storing an affiliation entry including an identifier of the affiliation entry and the list of affiliation nodes,the means for receiving (150) is further configured for receiving second FL task information including characteristics of a second requested FL training,the means for generating (170) is further configured for generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, andthe means for transmitting (160) is further configured for transmitting the second distributed nodes list, and whereingenerating the second distributed nodes list is further based on the stored affiliation entry.

3. The apparatus according to claim 2, whereinthe means for transmitting (160) is further configured for transmitting the identifier of the affiliation entry,the means for receiving (150) is further configured for receiving affiliation update information including an identifier of a corresponding affiliation entry, andthe storing means (180) is further configured for, in response to receiving the affiliation update information, updating the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation update information.

4. The apparatus according to claims 2 or 3, whereinthe means for transmitting (160) is further configured for transmitting the identifier of the affiliation entry,the means for receiving (150) is further configured for receiving affiliation delete information including an identifier of a corresponding affiliation entry, andthe storing means (180) is further configured for, in response to receiving the affiliation delete information, deleting the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation delete information.

5. The apparatus according to any one of claims 2 to 4, whereinthe storing means (180) is further configured for updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list in response to receiving the affiliation information.

6. The apparatus according to claim 5, whereinthe storing means (180) is further configured for updating the stored availability information corresponding to the affiliation nodes in response toreceiving the affiliation delete information and / or in response to receiving the affiliation update information.

7. The apparatus according to any one of claims 1 to 6, whereinthe storing means (180) is further configured for storing a list indicating reserved resources of distributed nodes in response to generating the distributed nodes list, the reserved resources of distributed nodes being reserved for the requested FL training;the means for receiving (150) is further configured for receiving second FL task information including characteristics of a second requested FL training,generating, in response to receiving the second FL task information, the second distributed nodes list is further based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes, andthe means for transmitting (160) is further configured for transmitting the second distributed nodes list, whereinresources of distributed nodes indicated in the list indicating reserved resources of distributed nodes are excluded from the second distributed nodes list.

8. The apparatus according to any one of claims 2 to 7, whereinthe received affiliation information further comprises a predicted FL task completion time,the storing means (180) is configured for storing the affiliation entry such that the affiliation entry further comprises storing the predicted FL task completion time, and whereinthe second distributed nodes list (S35) indicates a set of currently available distributed nodes and further indicates a set of currently unavailable distributed nodes together with an indication of a future time, when the currently unavailable distributed nodes will be available.

9. The apparatus according to any one of claims 2 to 8, whereinthe received affiliation information further comprises a first event exposure configuration,the second FL task information further comprises a second event exposure configuration,the means for generating is further configured for evaluating the first event exposure configuration and the second event exposure configuration, andthe means for generating is further configured for, in case the first event exposure configuration and the second event exposure configuration match, transmitting an event report and generating the second distributed nodes list only after the FL training process corresponding to the first event exposure configuration is finished.

10. The apparatus according to any one of claims 2 to 9, whereinthe received affiliation information further comprises FL process information about an FL training process,the storing means (180) is configured for storing the affiliation entry such that the affiliation entry further comprises storing the FL process information and an identifier of a corresponding FL aggregator,the means for receiving (150) is further configured for receiving affiliation retrieve information or second FL task information, the received information including second FL process information, the apparatus further comprisingmeans for searching for an affiliation entry with FL process information that fits the second FL process information of the received information, and whereinthe means for transmitting (160) is further configured for transmitting an affiliation retrieve response comprising the identifier of a corresponding FL aggregator included in the affiliation entry with FL process information that fits the second FL process information of the received information.

11. A method, comprising:receiving federated learning, FL, task information Including characteristics of a requested FL training (SI),generating, in response to receiving the FL task information, a distributed nodes list indicating, out of a plurality of distributed nodes, distributed nodes for contributing to the requested FL training (S2), andtransmitting the distributed nodes list (S4), whereingenerating the distributed nodes list is based on a comparison of the characteristics of the requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, and the method further comprisingin response to transmitting the distributed nodes list, updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list (S3).

12. The method according to claim 11, further comprising:receiving affiliation information comprising a list of affiliation nodes, indicating a subset of the set of distributed nodes indicated in the distributed nodes list (S6),creating an affiliation entry including an identifier of the affiliation entry and the list of affiliation nodes,storing the affiliation entry including the identifier of the affiliation entry and the list of affiliation nodes,receiving second FL task information including characteristics of a second requested FL training,generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes and on stored availability information corresponding to the plurality of distributed nodes, andtransmitting the second distributed nodes list, whereingenerating the second distributed nodes list is further based on the stored affiliation entry.

13. The method according to claim 11 or 12, further comprising:transmitting the identifier of the affiliation entry,receiving affiliation update information including an identifier of a corresponding affiliation entry, andin response to receiving the affiliation update information, updating the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation update information.

14. The method according to any one of claims 11 to 13, further comprising: transmitting the identifier of the affiliation entry, receiving affiliation delete information including an identifier of a corresponding affiliation entry, andin response to receiving the affiliation delete information, deleting the corresponding affiliation entry comprising the identifier of the affiliation entry included in the received affiliation delete information.

15. The method according to any one of claims 12 to 14, further comprising: updating the stored availability information corresponding to the set of distributed nodes indicated in the distributed nodes list in response to receiving the affiliation information.

16. The method according to any one of claims 13 to 15, further comprising: updating (S17) the stored availability information corresponding to the affiliation nodes in response to receiving the affiliation delete information and / or in response to receiving the affiliation update information.

17. The method according to any one of claims 11 to 16, further comprising: storing a list indicating reserved resources of distributed nodes in response to generating the distributed nodes list;receiving second FL task information including characteristics of a second requested FL training,generating, in response to receiving the second FL task information, a second distributed nodes list based on a comparison of the characteristics of the second requested FL training with stored characteristics of the plurality of distributed nodes, andtransmitting the second distributed nodes list, whereinresources of distributed nodes indicated in the list indicating reserved resources of distributed nodes are excluded from the second distributed nodes list.

18. The method according to any one of claims 12 to 17, whereinthe received affiliation information further comprises a predicted FL task completion time,storing the affiliation entry further comprises storing the predicted FL task completion time, and whereinthe second distributed nodes list (S35) indicates a set of currently available distributed nodes and further indicates a set of currently unavailable distributed nodes together with an indication of a future time, when the currently unavailable distributed nodes will be available.

19. An apparatus (200'), comprising:means for transmitting (260) federated learning, FL, task information including characteristics of a requested FL training,means for receiving (250) a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training,means for selecting (270) a subset of the set of distributed nodes, indicated by the received distributed nodes list, to be used for the requested FL training, whereinthe means for transmitting (260) is further configured for transmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes.

20. The apparatus according to claim 19, whereinthe means for receiving (250) is further configured for receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes,the means for transmitting (260) is further configured for transmitting, in response to the requested FL training being executed and completed, affiliation delete information including the received identifier of the affiliation entry.

21. The apparatus according to claim 19 or 20, whereinthe means for receiving (250) is further configured for receiving an identifier of an affiliation entry corresponding to the list of affiliation nodes, andthe means for transmitting (260) is further configured for transmitting, in response to starting a training iteration of the requested FL training, for which a part of the affiliation nodes are not used, affiliation update information including the received identifier of the affiliation entry and information about the affiliation nodes not used for the training iteration.

22. A method, comprising:transmitting federated learning, FL, task information including characteristics of a requested FL training,receiving a distributed nodes list indicating a set of distributed nodes suitable for contributing to the requested FL training,selecting a subset of the set of distributed nodes, indicated by the distributed nodes list, to be used for the requested FL training, andtransmitting, in response to selecting the subset of the distributed nodes, affiliation information comprising a list of affiliation nodes, indicating the selected subset of the set of distributed nodes.

23. A computer program comprising instructions, which when executed by an apparatus, cause the apparatus to perform the method according to any one of claims 11 to 18 or the method according to claim 22.

24. A non-transitory computer-readable medium comprising program instructions that, when executed by an apparatus, cause the apparatus to perform the method according to any one of claims 11 to 18 or the method according to claim 22.

25. A system, comprising:the apparatus according to any one of claims 1 to 10, andthe apparatus according to any one of claims 19 to 21.

Images