Dominik Klein

Installation without docker container

3

Hey, Great paper! Could you please provide instructions of how to install paddlepaddle without docker containers? Thanks a lot!

Negative loss in unbalanced Sinkhorn

3

Unbalanced Sinkhorn results in negative `reg_ot_cost`. ```python import jax from ott.core.sinkhorn import Sinkhorn from ott.geometry.pointcloud import PointCloud import jax.numpy as jnp from ott.core import LinearProblem n = 1000 dim =...

It would be great if we could access the data plotted in `cellrank.pl.aggregate_absorption_probabilities`, e.g. as a `pd.DataFrame`.

restructuring neural models + addition of OT-FM and GENOT

10

This is a PR for 1. new base classes neural solvers and models (i.e. neural networks) 2. Incorporating unbalancedness and learning the rescaling factors for any neural OT model. 3....

Throw an error if `scale_cost!=1.0` in `geometry.Geodesic`

2

**Is your feature request related to a problem? Please describe.** In `Geodesic`, the `scale_cost` argument has no effect. **Describe the solution you'd like** To prevent misunderstandings, I suggest to throw...

`scale_cost` does not work for kernels when it's not a `float`.

2

Whenever we have a kernel, the `scale_cost` argument does not work whenever it's not a `float` due to the `Geometry._cost_matrix` being `None` example which throws an error: ```python kernel =...

`annotate_compounds` for combinations of drugs

2

### Description of feature As a lazy but regular user of pertpy, would be great to have.

change `batch_size` in `sparsify`

11

Needed for keeping solver fast (high batch size), but prevent OOM in `sparsify` (need low batch size), see https://github.com/theislab/cellrank/issues/1146#issuecomment-1856213099

there are two options: - use callback for graph compputation (with epsilon=4t) - use `set_graph_xy`, `set_graph_x`, `set_graph_y`

Example on how to pass different callbacks in `prepare`

1

E.g. `pca-callback` and `spatial_norm_callback`.