It's a shame that the word categorical has already been applied to deep learning (categorical deep q networks a.k.a. C51, and Rainbow), because yours is arguably a more natural use of the name.