Closed plzhai closed 4 months ago
the reported baseline results in your paper are reproduced by yourselves?
they are also evaluated on all the locations of the dataset? I remember that GeoSAN evaluates their results on a sampled small set of all the locations.
hi, have you run STAN on other datasets and how did it work? thank you for reply
Hi, the idea in this article is somewhat interesting.
however, the running time is unacceptable absolutely, even on the provided toy data.
why using batch-size 1 and mask-len from 1 to len in the experiment? This setting helps better performance?
Thanks. Sorry if I did not check the repo regularly. The setting is not standard, and I personally do not recommend evaluating a POI model based on our own setting... The only reason we did so was precisely because of the limitations due to computational resources.
hi, have you run STAN on other datasets and how did it work? thank you for reply
hi, have you run STAN on other datasets and how did it work? thank you for reply
We only tested on the four datasets listed in the paper.
Hi, the idea in this article is somewhat interesting.
however, the running time is unacceptable absolutely, even on the provided toy data.
why using batch-size 1 and mask-len from 1 to len in the experiment? This setting helps better performance?