Fig. 4: Performance of co-training model.

The performance of SVR and KRR under a 3.0%, b 5.0%, c 10.0%, and d 15.0% labeling rates. e Comparison of model performance under different conditions at a 5.0% labeling rate. f The fitting performance of the ensemble model on the original label set. The fitting performance of g SVR and h KRR models on the final labeled set after co-training.