Fig. 2From: Assessing the effects of hyperparameters on knowledge graph embedding qualityBar charts showing Sobol sensitivities of the hyperparameters of knowledge graph embedding methods on three benchmark datasets. Note that dummy variables and continuous subparameters have been cumulatively grouped into their corresponding categories: training method, loss function, weight initialisation, and gradient descent algorithm. The raw data used to generate the figure are provided in Additional file 1: Table S3Back to article page