ToolsThe story, in brief
Using graph neural networks to recommend related products
160% improvement. That's what Amazon achieved using graph neural networks for product recommendations.
The KeyNews take
Why it matters
Amazon's research demonstrates significant performance gains in recommendation systems using advanced AI techniques, showing concrete metrics that other e-commerce and platform companies can benchmark against.
The key facts
4 to know30% to 160% improvements over predecessors
Dual embeddings approach
Graph neural networks for recommendations
Novel loss function implementation
Go to the source
Amazon Scienceamazon.science
Publisher excerpt: Dual embeddings of each node, as both source and target, and a novel loss function enable 30% to 160% improvements over predecessors.