Skip to main content
Fig. 4 | Applied Network Science

Fig. 4

From: SURREAL: Subgraph Robust Representation Learning

Fig. 4

PPI dataset: Comparison of embeddings per dimension for a random sample of 100 nodes. Node2vec, DeepWalk, Walklets, and SURREAL are run two times. The x-axis represents first run representations values, and the y-axis represents second run representations values. Three dimensions are selected randomly for each algorithm. The SURREAL-based representations are robust across runs (perfectly fall on a straight line y=x), which is not the case for node2vec, DeepWalk, and Walklets. The results are consistent for all the datasets. (a) Node2vec. Dimensions from left: 21, 48, 68 (b) DeepWalk. Dimensions from left: 5, 29, 120 (c) Walklets. Dimensions from left: 39, 55, 111 (d) SURREAL. Dimensions from left: 39, 55, 111

Back to article page