Antonis Klironomos defended his PhD

Antonis Klironomos has successfully defended his PhD thesis on “Democratization of Machine Learning using SemanticWeb Technologies”. In his thesis, he has analyzed how insights from a large number of past machine learning experiments, described in a rich knowledge graph, can be leveraged for determining suitable machine learning pipelines for a given prediction problem. To that end, he has developed a representation formalism for datasets and machine learning pipelines, methods utilizing knowledge graph embeddings for predicting the performance of a machine learning pipeline on a dataset, and a knowledge graph embedding method which can better exploit numerical values in a knowledge graph, which, this case, are of utmost importance, since the knowledge graph contains a lot of numeric knowledge like model hyperparameters and performance indicators.

The thesis was supervised by Heiko Paulheim at the University of Mannheim and Evgeny Kharalamov at BOSCH Research; the examination committee additionally comprised Christian Bizer and Frederik Armknecht.