About the dataset

[mdokamura-lmd]

Thank you for your impactful work.

MovieLens and LastFM datasets are used in the paper, but they appear to be different from those available on github.
For example, in the paper, the Training Records for MovieLens is 513,112, but in the published data, it is 1,000,209　(train/test=900,188/100,021). Furthermore, the ratio of the train/test is not 8:2.

Is the published dataset the one used for the experiments in the paper?

Thanks.