Pre-trained Machine Learning models are integral to modern artificial intelligence applications but suffer from a lack of standardization and accessibility. This thesis introduces an ontology-based framework to categorize, search, and retrieve pre-trained ML models.
The ontology captures key attributes such as architecture, training data, performance metrics, and application domains. A proof-of-concept web repository, populated with models from various sources, demonstrates the practical implementation of the ontology. This ontology-based approach addresses the need for standardization and paves the way for more efficient utilization of pre-trained ML models.