search

LukasZahradnik / deep-db-learning

A modular message-passing scheme reflecting the relational model for end-to-end deep learning from databases
https://lukaszahradnik.github.io/deep-db-learning/
6 stars 2 forks source link

Strange behavior of BFS strategy #19

Closed neumannjan closed 1 year ago

neumannjan commented 1 year ago

Try the mutagenesis dataset with the BFS strategy on depth 4 and with other depths. For me, I get twice as many "atom" entries with depth 4. Seems that each atom is repeated twice when depth is set to 4, with sometimes there being a "None" value and sometimes a non-None value in the 2nd column. Strange.

neumannjan commented 1 year ago

This also happens with (some?) other datasets at different depths

LukasZahradnik commented 1 year ago

Good catch. This was an oversight from a previous implementation where multiple samples were processed at once (now there is "get" for only one sample). Fixed