Fuli Feng
Fuli Feng
Yes, this is not a good way to construct label. But this is a hard choice. During my rejected submission, there are comments like "Should follow the same setting as...
Did you try to uncompress the relation.tar.gz?
According to the released sequential embeddings and the default parameters, it should be easy to figure out the correct dimensions.
No, I didn't try that.
Come on! Please check the difference when we calculate: pre_pw_dif and gt_pw_dif
You can find it through the link in README! To repeat the experiment, i.e., train a RSR model, downloaded the pretrained sequential embedding, and extract the file into the data...
Yes, before we use the maximum value throughout the entire 2013-2017 dataset, we have tried to use the maximum value during the training period. The results are quite close.
Did you try to uncompress the relation.tar.gz?
No, -1234 is just a special value which means the data is missing. You can replace it with any negative number that is large enough.
It's not an issue. We subtract 1.0 before report performance. The consideration of not initializing it with 0.0 is the potential of running back-testing where the total money changes along...