Question: What is the difference between Link prediction pipelines and the typical GDS link prediction algorithms like Adar? Is there a way to combine them?
The link prediction algorithms in the GDS are self-contained algorithms, which define some closeness metric with the goal to predict links. The major difference to a link prediction pipeline (which has the same objective) is, that the pipeline is based on a machine learning approach. Meaning that the pipeline is based on a predefined feature set, splits the data into train and test data and then trains a statistical model (based on multiple features) which is then in turn used to predict links on unseen data. In a sense, the pipeline approach tries to find a generalizable model which can be applied to unseen data (which is not the case for the link prediction algorithms). Hope this helps.
Question: What is the difference between Link prediction pipelines and the typical GDS link prediction algorithms like Adar?
Is there a way to combine them?
The link prediction algorithms in the GDS are self-contained algorithms, which define some closeness metric with the goal to predict links. The major difference to a link prediction pipeline (which has the same objective) is, that the pipeline is based on a machine learning approach. Meaning that the pipeline is based on a predefined feature set, splits the data into train and test data and then trains a statistical model (based on multiple features) which is then in turn used to predict links on unseen data. In a sense, the pipeline approach tries to find a generalizable model which can be applied to unseen data (which is not the case for the link prediction algorithms).
Hope this helps.
@philippbrunenberg yes most helpful! Tq