InstantEmbedding: Efficient Local Node Representations

TL;DR

InstantEmbedding is a novel, highly efficient method for generating local node representations using PageRank, achieving significant speed and memory improvements while maintaining or surpassing state-of-the-art quality in graph tasks.

Contribution

The paper introduces InstantEmbedding, a new local node embedding technique that is theoretically sound and empirically vastly more efficient than existing methods.

Findings

Over 9,000 times faster computation time.

Over 8,000 times less memory usage.

Matches or exceeds state-of-the-art performance in node classification and link prediction.

Abstract

In this paper, we introduce InstantEmbedding, an efficient method for generating single-node representations using local PageRank computations. We theoretically prove that our approach produces globally consistent representations in sublinear time. We demonstrate this empirically by conducting extensive experiments on real-world datasets with over a billion edges. Our experiments confirm that InstantEmbedding requires drastically less computation time (over 9,000 times faster) and less memory (by over 8,000 times) to produce a single node's embedding than traditional methods including DeepWalk, node2vec, VERSE, and FastRP. We also show that our method produces high quality representations, demonstrating results that meet or exceed the state of the art for unsupervised representation learning on tasks like node classification and link prediction.