How to prevent overfitting in supervised mode?

Nobody has claimed this yet.

Using standard parameters in supervised mode with dichotomous response variable (0/1) and embedding in 2 dimensions, I fit on 80% of data (800k points,20 cols) and I transform on 80% and on the holdout 20% of data.

This are the results I'm getting:


As you can see for training data I get a clear separation between zeros and ones but in the test dataset no clear picture is produced. This is clearly an evidence of overfitting (for reference fitting xgboost on the same dataset yields an AUC of 0.8).

Which are the parameters I can tweak in order to avoid this effect?

Thanks for the support,
Michele.

Start by reviewing the supervised-mode example described in the issue, including the 80/20 split, dichotomous response, and 2-dimensional embedding. No source files or tests are named, so identify the supervised fitting entry point and existing parameter documentation first; done would require a justified way to reduce the reported train/holdout overfitting and documentation or tests showing the behavior.

Written by the indexing model from the issue text.