How to normalize scores as described in "An Analysis of Fusion Functions for Hybrid Retrieval"

How do I normalize retrieval scores as described in ["An Analysis of Fusion Functions for Hybrid Retrieval"](https://arxiv.org/abs/2210.11934)

Here is an example of several normalization functions:
image

I’d be most interested in using T_MM on the sparse vector score, as this is what the authors use. Given that the paper is authored by folks at Pinecone, I’m hoping this is already possible.