| CPC G06N 3/082 (2013.01) [G06F 17/16 (2013.01); G06F 18/2148 (2023.01); G06N 3/047 (2023.01); G06N 3/08 (2013.01); G06N 5/02 (2013.01)] | 18 Claims |
|
1. A computer-implemented method for training a graph neural network (GNN) model based on an entity-entity graph for relationship prediction, the entity-entity graph comprising a plurality of entity nodes in which each entity node is connected to one or more entity nodes of the plurality of entity nodes by one or more corresponding relationship edges, the method comprising:
generating an embedding based on data representative of at least a portion of the entity-entity graph for the GNN model, wherein the embedding comprises an attention weight assigned to each relationship edge of at least the portion of the entity-entity graph; and
updating the GNN model including the attention weights by minimising a loss function
 (⋅) associated with at least the embedding;wherein the attention weights indicate the relevancy of each corresponding relationship edge between entity nodes of the entity-entity graph and assist the GNN model in relationship prediction,
wherein minimising the loss function,
(∩), comprises:assigning an attention weight, αr,i,j, to each relationship edge of the entity-entity graph, wherein αr,i,j, is the attention weight associated with the connection between entity node i and entity node j with relation index r of the plurality of relationships associated with the entity-entity graph;
modifying the loss function by adding an attention loss function,
att(Λ) associated with the set of attention weights Λ={αr,i,j} to the loss function; andminimising the modified loss function,
mod, with respect to the weights of the GNN model including the attention weights;wherein the modified loss function is
mod(⋅)=(⋅)+att(Λ). |