Check for singularity?

[Weekend-Warrior]

Does it make sense to check for and remove linear combinations in the expanded matrix? Scenarios such as the coincidence of missing values should probably be condensed to a single column. In some cases I don't think it makes a difference but if the modeling algorithm relies on non-singularity or incorporates randomization of covariates, it should have an impact.

[liuauto]

Hi,
I am thinking the same ... Hope this can be added as a new feature. Can we add the option for categorical feature with n levels, only generating n -1 features instead of n features. In some scenario, this could be very useful.