Recursive Learning Framework for Structured Data Agglomeration
Angelo Ciaramella, Emanuel Di Nardo, Giuseppe Vettigli · 2024
In the current era, vast amounts of data are readily available, particularly in the form of multi-structured data such as sequences, trees, and graphs. Analyzing these diverse data types requires specialized approaches. However, standard machine learning algorithms are not always suitable due to their lack of adaptability to the inherent nature of structured data. To address this limitation, our work uses a novel learning framework based on a general recursive scheme. This framework effectively embeds structured data into vector representations and leverages Principal Component Analysis to create a clustering technique. We conduct comparisons and experiments with established algorithms to evaluate performance across synthetic and real-world datasets.