This is the corrected version for Self-Attentive Hawkes Process (SAHP).
We have discussed with the first author of the SAHP paper, Qiang Zhang for the confirmation and agreement of the corrections.
-
For training, please go into
sahp_training
sub-package and follow the instructions there. You will obtain the modelmodel.pt
that can be used for testing purposes. Please note that the RMSE and Accuracy stuffs in the training log is not reliable. For evaluation purposes (computing loglik, RMSE, Accuracy), please usesahp_testing
sub-package, which will be introduced below. -
For testing and synthetic data drawing, as the SAHP authors do not provide thinning algorithm implementation, for a fair comparison, we create a separate sub-package
sahp_testing
to do that.