iN2V: Bringing Transductive Node Embeddings to Inductive Graphs

Nicolas Lell, Ansgar Scherp

TL;DR: We introduce iN2V, a method extending Node2Vec to generate embeddings for unseen nodes in the inductive setting.

Abstract:

Shallow node embeddings like node2vec (N2V) can be used for nodes without features or to supplement existing features with structure-based information.
Embedding methods like N2V are limited in their application on new nodes, which restricts them to the transductive setting where the entire graph, including the test nodes, is available during training.
We propose inductive node2vec (iN2V), which combines a post-hoc procedure to compute embeddings for nodes unseen during training and modifications to the original N2V training procedure to prepare the embeddings for this post-hoc procedure.
We conduct experiments on several benchmark datasets and demonstrate that iN2V is an effective approach to bringing transductive embeddings to an inductive setting.
Using iN2V embeddings improves node classification by 1 point on average, with up to 6 points of improvement depending on the dataset and the number of unseen nodes.
Our iN2V is a plug-in approach to create new or enrich existing embeddings.
It can also be combined with other embedding methods, making it a versatile approach for inductive node representation learning.
Code to reproduce the results is available at https://github.com/Foisunt/iN2V.

Lay Summary:

In real-world situations, graphs (like social networks) often change over time — for example, when new users create accounts.
These changes add new nodes and connections that weren’t available during the original training of the embedding model.
Many popular graph embedding methods, such as Node2Vec, can only create embeddings for nodes that were known during training.
To handle new, unseen nodes, we introduce a method that can create embeddings for them after the original training is done.
Although our method can work with different embedding methods, we focus on Node2Vec and adjust its training so it produces embeddings that are easier to extend later.
This approach leads to better performance in real-world tasks where it's unrealistic to assume all nodes are known in advance.

Link To Code: https://github.com/Foisunt/iN2V

Primary Area: Deep Learning->Graph Neural Networks

Keywords: Node2Vec, Inductive, Embedding, Graph, GNN

Submission Number: 11140