From random-walks to graph-sprints: a low-latency node embedding framework on continuous-time dynamic graphs

TL;DR

This paper introduces graph-sprints, a low-latency, streaming framework for node embedding in continuous-time dynamic graphs, enabling real-time inference with competitive accuracy and significantly reduced latency.

Contribution

The paper presents a novel streaming approximation method for node embeddings in CTDGs, suitable for real-time applications, outperforming existing high-latency models.

Findings

Achieves near an order of magnitude speed-up in inference latency.

Outperforms state-of-the-art algorithms in node classification accuracy.

Demonstrates effectiveness across multiple datasets.

Abstract

Many real-world datasets have an underlying dynamic graph structure, where entities and their interactions evolve over time. Machine learning models should consider these dynamics in order to harness their full potential in downstream tasks. Previous approaches for graph representation learning have focused on either sampling k-hop neighborhoods, akin to breadth-first search, or random walks, akin to depth-first search. However, these methods are computationally expensive and unsuitable for real-time, low-latency inference on dynamic graphs. To overcome these limitations, we propose graph-sprints a general purpose feature extraction framework for continuous-time-dynamic-graphs (CTDGs) that has low latency and is competitive with state-of-the-art, higher latency models. To achieve this, a streaming, low latency approximation to the random-walk based features is proposed. In our framework, time-aware node embeddings summarizing multi-hop information are computed using only single-hop operations on the incoming edges. We evaluate our proposed approach on three open-source datasets and two in-house datasets, and compare with three state-of-the-art algorithms (TGN-attn, TGN-ID, Jodie). We demonstrate that our graph-sprints features, combined with a machine learning classifier, achieve competitive performance (outperforming all baselines for the node classification tasks in five datasets). Simultaneously, graph-sprints significantly reduce inference latencies, achieving close to an order of magnitude speed-up in our experimental setting.