Dominik Klein

Examples to be written

5

- [x] passing custom cost matrices - [ ] `LineageProblem`: passing barcodes - [ ] `problem_stage` - [ ] `plot_cell_transition`

To prevent low-quality results when plotting the interpolated colors in a Sankey diagram we should make the "stepsize" a variable (as opposed to 50) [here](https://github.com/theislab/moscot/blob/0a83ede4633be07859ab6b4c7336286a22581fc4/src/moscot/plotting/_utils.py#L169)

Currently, we get different results when doing this because of the `normalize` argument, which by default is set to `True`. Intuitively, when using push/pull on either of the two options,...

remove generic types

1

for the sake of readability of the docs.

I want to train on a subset of control cells, and perform prediction on another set of control cells. how can i do that?

Currently, we don't document that you can also pass a full-rank initializer. Maybe it's redundant, but this is to be discussed.

adapt low rank gamma default for linear problems

1

for GW we keep default of 10.0, but for linear LR we need to set it to 500.