Self-labeling Sample Clauses

Self-labeling. We evaluated the usefulness of the self-labeling component by showing how the test accuracy evolves after each co-training iteration. Figure 5 shows that the accuracy generally has an increasing trend with more co-training iterations. In some cases, the final iterations may have a decreasing trend, because in the last few iterations the model self-labels the samples that it is most uncertain about, and thus it is more likely to make mistakes. For this reason, we kept track of the validation accuracy, and at the end we restored the model from the co-train iteration with the best validation accuracy. Self-labeling is also a critical component for datasets such as Pubmed, where in the first co-train iteration there are no edges with both nodes labeled, so g cannot be trained until we self-label more nodes. In such cases, g returns 1 by default until it can be trained, defaulting to NGM and relying on the graph (although for noisy graphs, one could return 0 by default).