Heterogeneous tree graph neural network for label prediction

A system comprising:

• a processor;
• memory;
• a label predictor stored in the memory and executable by the processor to:
• receive as input a heterogeneous graph structure that represents a real-world network;
• identify, for each node of multiple nodes in the heterogeneous graph structure, a set of node-target paths that connect the node to a target node;
• assign, to each path of the node-target paths identified for each node, a path type identifier indicative of a number of edges and corresponding edge types in the path;
• extract a semantic tree from the heterogeneous graph structure, the semantic tree comprising the target node as a root node and defining a hierarchy of metapaths that each individually correspond to a subset of the node-target paths in the heterogeneous graph structure assigned to a same path type identifier;
• encode, with one or more neural networks, a set of metapath embeddings corresponding to nodes of the semantic tree, each of the metapath embeddings comprising aggregated feature-label data for nodes in the heterogeneous graph structure corresponding to the path type identifier associated with the metapath represented by the metapath embedding;
• and
• generate a label prediction for the target node in the heterogeneous graph structure based on the set of metapath embeddings.

A tangible computer-readable storage media encoding computer-executable instructions for executing a computer process, the computer process comprising:

• determining a heterogeneous graph structure representing a real-world network;
• identifying, for each node of multiple nodes in the heterogeneous graph structure, a set of node-target paths that connect the node to a target node;
• assigning, to each path of the node-target paths, a path type identifier indicative of a number of edges and corresponding edge types in the path;
• extracting a semantic tree from the heterogeneous graph structure, the semantic tree comprising the target node as a root node and defining a hierarchy of metapaths that each individually correspond to a subset of the node-target paths in the heterogeneous graph structure assigned to a same path type identifier;
• encoding, with one or more neural networks, a set of metapath embeddings corresponding to nodes of the semantic tree, each of the metapath embeddings comprising aggregated feature-label data for nodes in the heterogeneous graph structure corresponding to the path type identifier associated with the metapath represented by the metapath embedding;
• and
• predicting a label of the target node in the heterogeneous graph structure based on the set of metapath embeddings.