Machine learning model

Machine learning model: a software artifact produced from a training process. This algorithmic structure conducts inference or prediction operations, based on the provided input data. The model processes the input data, based on the learned rules, captured patterns from training data. In the context of LLM’s, this asset processes textual input data and produces text output. The “1” notation, applied to the associations at the model’s side indicates that the associated assets apply only to 1 specific model in 1 instance.

The model has a defined attribute “ML model state”, which is an enumeration, which can contain a value from the predetermined set of values. The values are – Preprocessed, Trained, Refined and Deployment. This attribute describes the state of the target model, as it processed through the defined process pipelines. Initially, the model is designed with a set architecture and a set of parameters, untrained – this is the Preprocessed state, an initial design and an empty model. Afterwards, the Preprocessed model is trained on the target datasets, producing an intermediary state – Trained, when the model has adjusted parameters, but the training is not yet finalized. The model is further trained, tested and refined, producing a final model, suitable for use in a production environment, in this case the model is in the Refined state. Finally, the refined model is deployed to the target environment and is used for conducting necessary operations, this is the Deployed state.

Relationship with component business assets:

  1. Model parameters: integral components of a model, internal variables, critical to its operation and utility. In the context of the work, the following components are considered as model parameters: layers, type of activation function, layer connections, parameters, weights. The values of parameters are adjusted during the training
  2. Intellectual property: legal rights and protections, proprietary methods, training methodologies, curated training data and designs associated with the produced machine learning model and its components. The intellectual property underlines a company’s competitive advantage, it can incorporate developed trade secrets and patents. The “1*” notation indicates that a model may “have one intellectual property”.