Adaptive model partitioning

• Optimize the partition routing method of the data server model, which can support the partition of range and hash at the same time. In the actual graph algorithm training process, the appropriate model partition method can be adaptively selected according to the calculation characteristics of different algorithms

Complex heterogeneous Graph Embedding

• Enriched and expanded the storage structure and calculation mode of the graph, and provided a flexible custom ps func interface for complex operations, which can well support the storage and calculation of complex heterogeneous graph networks, and can support high-dimensional sparse graph node features. Easily perform representation learning of heterogeneous graphs
• Implemented 5 out-of-the-box heterogeneous graph neural network algorithms: HAN, heterogeneous GAT, heterogeneous GraphSage, IGMC edge prediction, and heterogeneous Bipartite GraphSage

High-performance optimization of large graphs with hundreds of billions of edges

• Special performance optimizations have been made for the training of hundreds of billions of edges, and the test results of the k-core and common friends algorithm are provided.

Enrich the content of machine learning algorithm library

• Added more than a dozen feature engineering methods such as Correlation, Discrete,MutualInformationRandomizedSVD, etc., and a multi-task learning algorithm esmm