The use of standardized evaluation procedures is a key component in the Machine Learning (ML) field to determine whether new approaches grant real advantages over others. This is especially true for fast-growing research areas, where a substantial amount of literature relentlessly appears every day. In the graph machine learning field, some evaluation issues have already been brought to light and partially addressed, but a general-purpose library for rigorous evaluations and reproducible experiments is lacking. We therefore introduce a new Python library, called PyDGN, to provide users with a system that lets them focus on models’ development while ensuring empirical rigor and reproducibility of their results.
PyDGN: a Python Library for Flexible and Reproducible Research on Deep Learning for Graphs
Federico Errica;Davide Bacciu;Alessio Micheli
2023-01-01
Abstract
The use of standardized evaluation procedures is a key component in the Machine Learning (ML) field to determine whether new approaches grant real advantages over others. This is especially true for fast-growing research areas, where a substantial amount of literature relentlessly appears every day. In the graph machine learning field, some evaluation issues have already been brought to light and partially addressed, but a general-purpose library for rigorous evaluations and reproducible experiments is lacking. We therefore introduce a new Python library, called PyDGN, to provide users with a system that lets them focus on models’ development while ensuring empirical rigor and reproducibility of their results.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.