US Patent Application 18325348. NEURAL NETWORK TRAINING AND INFERENCE WITH HIERARCHICAL ADJACENCY MATRIX simplified abstract
Contents
NEURAL NETWORK TRAINING AND INFERENCE WITH HIERARCHICAL ADJACENCY MATRIX
Organization Name
Inventor(s)
Emmanouil Ioannis Farsarakis of Livingston (GB)
Michal Krzysztof Szarmach of Gdynia (PL)
NEURAL NETWORK TRAINING AND INFERENCE WITH HIERARCHICAL ADJACENCY MATRIX - A simplified explanation of the abstract
This abstract first appeared for US patent application 18325348 titled 'NEURAL NETWORK TRAINING AND INFERENCE WITH HIERARCHICAL ADJACENCY MATRIX
Simplified Explanation
- This patent application describes a method for generating hierarchical adjacency matrices of a graph for deep neural network (DNN) training or inference. - The graph consists of nodes that are connected by edges. - One or more target nodes are selected from the graph. - A hierarchical sequence of node groups is formed, where each node group represents a neighborhood in the graph. - The first node group includes the target node(s). - Subsequent node groups include nodes directly connected to any node in the previous node group. - A hierarchical adjacency matrix is generated based on the hierarchical sequence. - The hierarchical adjacency matrix represents the connections between nodes in the node groups. - Each row in the matrix represents a node in the graph. - The rows are arranged in accordance with the hierarchical sequence. - The elements in the matrix encode the edges between the nodes in the node groups.
Original Abstract Submitted
Hierarchical adjacency matrices of a graph may be generated for DNN training or inference. The graph includes nodes connected by edges. One or more target nodes may be selected from the graph. A hierarchical sequence of node groups may be formed. A node group may be a neighborhood in the graph. A first node group (e.g., 0-hop neighborhood) may include the target node(s). A subsequent node group (e.g., 1-hop neighborhood, 2-hop neighborhood, etc.) may include one or more nodes directly connected to any node of the previous node group in the hierarchical sequence. A hierarchical adjacency matrix may be generated based on the hierarchical sequence. The hierarchical adjacency matrix may include rows, each of which rows represents a respective node in the graph. The rows may be arranged in accordance with the hierarchical sequence. The hierarchical adjacency matrix may include elements encoding edges between the nodes in the node groups.
