worldbank / REaLTabFormer

A suite of auto-regressive and Seq2Seq (sequence-to-sequence) transformer models for tabular and relational synthetic data generation.
https://worldbank.github.io/REaLTabFormer/
MIT License
207 stars 23 forks source link

Question for Generate Synthetic data To compare HMA and REaLTabFormer #67

Open limhasic opened 6 months ago

limhasic commented 6 months ago

Aren't there explicit tables among the tables?

For example, in the case of hotel-guest data to test the HMA algorithm in SDV, when the hotel data is explicitly used, how can REaLTabFormer synthesize only the guest data while leaving the hotel data as is?

Should hotel-guest data be viewed as a secondary format?

avsolatorio commented 5 months ago

Hello @limhasic , sorry but I am not familiar with the hotel-guest data. But based on what I understand, you should be able to use the relational mode to train a model that produces synthetic guest data based on the hotel data as input.

limhasic commented 5 months ago

image

like this data If hotel becomes the parent table in this data, shouldn't synthesis be done only in guest?