| CPC H04L 41/0895 (2022.05) [H04L 41/0873 (2013.01); H04L 41/16 (2013.01)] | 1 Claim |
|
1. An apparatus for implementing a training coordination function, the apparatus comprising at least one processor and at least one memory including computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus at least to perform:
processing at least one training request with which training of a network function instance of an autonomous network of a communication network system is requested;
based on a result of the processing, deciding to approve the at least one training request; and
based on the deciding to approve the at least one training request,
planning the training based on the at least one training request;
returning an identification of the at least one training request to the network function instance; and
in accordance with the planning, storing or updating an existing training state associated with the network function instance in a database which is configured to store existing training states associated with network function instances of the autonomous network;
retrieving, from the database, one or more existing training states associated with the at least one training request,
wherein the processing comprises: processing the one or more existing training states and the at least one training request;
combining metadata of a network function type corresponding to the network function instance with data of a training descriptor included in the at least one training request, to thereby create metadata of the network function instance, wherein the metadata of the network function type has been provided at a design time of the training coordination function;
using the metadata of the network function instance to perform the following:
retrieving the one or more existing training states from the database;
the processing the at least one training request; and
processing the one or more existing training states and the processing the at least one training request;
providing training state information including recommendations for retraining with new available training data or reuse of a trained model to the network function instance; and
acquiring updates on events relevant to the training and updating the one or more existing training states in the database based on these events;
the training descriptor includes:
information about the network function type,
information about a network function version,
information about an algorithm type,
information about an algorithm version,
information about the network function instance,
information about a spatial scope associated with the network function instance,
information about a temporal scope associated with the training the network function instance,
information about constraints associated with data used in the training,
information about a type of trigger that has triggered the at least one training request,
information about a type of reason that has triggered the at least one training request, and
information about a reference to an earlier training state via a training request identification.
|