A "hidden" feature of tabular foundation models
Don’t we have quantile regression though?
We do. Depending on the approach, it requires either having multiple models (like 3 separate xgboost models I needed in my example) or having strong assumptions on the distribution, then you can use, e.g., Bayesian approaches.
Don’t we have quantile regression though?
We do. Depending on the approach, it requires either having multiple models (like 3 separate xgboost models I needed in my example) or having strong assumptions on the distribution, then you can use, e.g., Bayesian approaches.