Thesis Degree: Master Thesis in Computer Science
In collaboration with the Model-Driven Engineering Group: Prof. Dr. Davide Di Ruscio, Dr. Phuong Nguyen, and Dott. Claudio Di Sipio.
In Model-Driven Engineering, metamodels are used for formalizing notations used by experts for expressing problems in a given domain. As a consequence, metamodels are central components of complex ecosystems consisting of numerous metamodel-dependent artifacts, including editors, transformation and analysis programs, and code generators. Similar to other software artifacts, metamodels are prone to evolution. Whenever a metamodel undergoes a modification, the validity of the other components of the ecosystem might be compromised. Therefore, it is important to have techniques and methods that, starting from the metamodel changes it can drive the (semi) automated adaptation of the ecosystem.The recent development of several disruptive Machine Learning (ML) algorithms holds promise for success in different application domains. Machine Learning algorithms are capable of simulating humans' learning activities, aiming to acquire real-world knowledge autonomously. In this way, ML systems can generalize from concrete examples, identify patterns and make decisions by means of data, without being manually coded or intervened by humans. Thanks to this characteristic, ML algorithms have been widely applied in several domains, such as Web search, recommender systems, self-driving cars, to name a few.This thesis intends to investigate advanced Machine Learning techniques for providing adaptation procedure capable of restoring the validity of models, transformations, and the other artifacts in the ecosystem affected by the metamodel evolution.
Prof. Alfonso Pierantonio