Supplementary material for Gaussian Copula Embeddings 1 General settings

Our model is implemented in the R language. During the computation, no parallel computation nor GPU resources are needed. In this section we provide more details related to the Section 5 in the main paper. As mentioned in the main paper, We use 10% held-out data as the testing data set and rest 90% of the data as the training set. MAE and standard error (in the parentheses) can be found in Table 2. Results corresponding to Section 5.1, 5.2 and 5.4 in the main paper can be found in Table 4. Results corresponding to Section 5.3 can be found in Table 5.