A Appendix A.1 Additional Experiments

We analyze how our MeT A and DA techniques work for temporal graph learning. Next, we provide therotical analysis to show that our methods meet this expectation. Based on Alg. 1, because edges This theorem guarantees that our DA techiniques do not break the original edge time distribution. Otherwise, if the original distribution is broken, e.g., all the edge time decrease to 0, the augmented edges cannot reflect the practical condition and may degrade the models' generalization. DA technique does not change the number of edges.