AI Insight
Researchers developed FLASH, a new adaptive method for sampling historical data in temporal graph neural networks that learn social network patterns over time. Unlike traditional fixed sampling approaches that either select random past interactions or only recent ones, FLASH dynamically adjusts which historical information to examine based on the specific structure of each network. Testing across multiple datasets showed that temporal graph neural networks using FLASH consistently outperformed those using conventional sampling methods.
Why it matters
This advancement could make AI systems that analyze evolving social networks, recommendation systems, and communication patterns significantly more efficient and accurate while using fewer computational resources. The approach has practical applications in predicting future connections in social media, fraud detection, and modeling how information spreads through networks.
Understand the Science
arXiv:2504.07337v2 Announce Type: replace
Abstract: Aggregating temporal signals from historic interactions is a key step in future link prediction on dynamic graphs. However, incorporating long histories is resource-intensive. Hence, temporal graph neural networks (TGNNs) often rely on historical neighbors sampling heuristics such as uniform sampling or recent neighbors selection. These heuristics are static and fail to adapt to the underlying graph structure. We introduce FLASH, a learnable and graph-adaptive neighborhood selection mechanism that generalizes existing heuristics. FLASH integrates seamlessly into TGNNs and is trained end-to-end using a self-supervised ranking loss. We provide theoretical evidence that commonly used heuristics hinder TGNNs performance, motivating our design. Extensive experiments across multiple benchmarks demonstrate consistent and significant performance improvements for TGNNs equipped with FLASH.
Source: FLASH: Flexible Learning of Adaptive Sampling from History in Temporal Graph Neural Networks