Large Scale Graph Representation in Application at Alibaba
Large-scale graph representation is playing an increasingly important role for big data companies. In particular, graph inference combined with deep learning has achieved successful phased results in many of Alibaba’s business scenarios. The data of the Alibaba ecosystem is extremely rich and varied, covering everything from shopping, travel, and entertainment to payments.
A specific example of this technology is our personalized e-commerce recommendation system, which filters information and recommends suitable products or services to consumers based on their habits and hobbies. Traditional recommendation systems are prone to issues with sparsity, cold starts and information repeatability. Large-scale graph representation is emerging as a recommended source of auxiliary information for effectively collating data from global consumers. Utilizing this technology can help companies truly understand the needs of their consumer base and create products that engage consumers across a variety of business domains.
Alibaba is currently developing a new generation of graph learning platforms capable of efficiently performing inference analysis on billions of nodes and trillions of edges. Several of our recent KDD-accepted academic papers on this topic have received recognition, including our papers on fraud detection applications (SPARC: Self-Paced Network Representation for Few-Shot Rare Category Characterization), recommendation systems (Interactive Paths Embedding for Semantic Proximity Search on Heterogeneous Graphs), entity resolution (Mobile Access Record Resolution on Large-Scale Identifier-Linkage Graphs), and deep model interpretation (Adversarial Detection with Model Interpretation).
To emphasize the importance of this technology in our future endeavors, Alibaba is also setting up a graph computing lab at Alibaba’s DAMO academy.
If you are interested in learning more about Alibaba’s large-scale graph representation technology, click here.