07 August 2024
In this work, EMERGE partners from the University of Pisa investigate neural models based on graph random features for classification tasks. First, the authors aim to understand when over parameterization, namely generating more features than the ones necessary to interpolate, may be beneficial for the generalization abilities of the resulting models.
They employ two measures: one from the algorithmic stability framework and another one based on information theory. They provide empirical evidence from several commonly adopted graph datasets showing that the considered measures, even without considering task labels, can be effective for this purpose. Additionally, they investigate whether these measures can aid in the process of hyperparameters selection. The results of their empirical analysis show that the considered measures have good correlations with the estimated generalization performance of the models with different hyperparameter configurations and can be used to identify good hyperparameters, achieving results comparable to the ones obtained with a classic grid search.
Read the paper in the link below.
More Information
