search

sattarov / FinDiff

Implementation of the paper: "FinDiff: Diffusion Models for Financial Tabular Data Generation"
19 stars 2 forks source link

Help #2

Open JiangLei1012 opened 2 months ago

JiangLei1012 commented 2 months ago

Hello sir，this is an excellent project, thank you for your paper.

I'm not smart enough to know your code particularly well. Could you please give me a more detailed explanation? Thank you so much!

Best wish, Chris

JiangLei1012 commented 2 months ago

If I understand you correctly, we binary classify the dataset and the classification feature is thecategorical attribute, i.e. we are classifying based on the CATEGORICAL attributes. So what are the numeric attributes used for?

sattarov commented 1 month ago

Hi Chris,

I hope I understood your question correctly. In case the label is not provided we pick a categorical attribute as a label. The numeric attributes are part of the regular training set, together with other categorical attributes. The selection of the label is specific to your use case and is not mandatory, the model can also be trained without a label (unconditional training/sampling). However, to measure the Utility of the model, we need labels. Hope that helps.

Best regards, Timur