Parameter Tuning Sample Clauses

Parameter Tuning. Table 4.2 shows the parameter settings with default values highlighted with an un- derline for Gowalla dataset and in bold font for California dataset. We use the grid granularity gQP = 512 to compute the query plan in all of the methods. We tune the grid granularity for the indexing and query processing and other parameters for each method as follows. Table 4.2: Parameter settings Parameter Values gn (Naive Grid Granularity) 32 64 128 256 g (RkNN-HG Grid Granularity) 32 64 128 256 512 h (RkNN-HRT Xxxxxxx Resolution) 32 64 128 256 512 1024 b (Node Block Size (B)) 32 64 128 256 512 t (MBR Size Threshold) 32 64 128 256 512 gQP (Query Plan Grid Granularity) 512 Computational Cost (sec) Overhead Size (Relative) Computational Cost (sec) 25000 20000 15000 10000 5000 0 32 64 128 Grid Granularity (g) 16 Overhead Size (relative) 14 10 8 6 4 2 0 256

Parameter Tuning. SVNE (and MVNE) are compared with other single view methods (and their multi-view extensions) using the code provided by the authors of the respective methods (with the relevant parameters set or tuned as speci- fied in the respective papers). We explored several different settings for d, the dimension of the embedding space (64, 128, 256, 512) for all the methods. We used grid search over 1xxxxx://xxx.xxxx.xx/ 2xxxxx://xxx.xxxxxx.xxx/ γ ∈ {40, 80} for Deepwalk and p, q ∈ {0.25, 0.50, 1, 2, 4} for node2vec. Performance Evaluation: In experiments 1-2, we measure the performance on the node label prediction task using different fractions of the available data (10% to 90% in increments of 10%) for training and the remaining for testing the predictors. In experiment 3, we use 50% of the nodes in each view for training and the rest for testing. We repeat this procedure 10 times, and report the performance (as measured by Micro F1 and Macro F1) averaged across the 10 runs. In each case, the embeddings are evaluated with respect to the performance of a standard one-versus-rest L2-regularized sparse logistic regression classifiers [14] trained to perform node label prediction.