iiinova — where your digital self keeps evolving

Sebastian Bischoff

@sbischoff.bsky.social

122 followers 104 following 8 posts

PhD student @mackelab.bsky.social and claassenlab in Tübingen

Sebastian Bischoff @sbischoff.bsky.social · 05/11/2025

In some sense, that's a weaker requirement than to also test how plausible the embedding of x+\Delta t*v is. I could imagine that this leads to problems for a stationary simulation.

100

Sebastian Bischoff @sbischoff.bsky.social · 05/11/2025

Were you thinking of a specific simulation? The steady-state bursty model in your paper? The test doesn't consider the magnitude of the low-d or high-d velocity vector, just the direction. We decided to do it this way because of the various scaling issues of the velocity estimate.

200

Sebastian Bischoff @sbischoff.bsky.social · 04/11/2025

This is an effect of the test statistic being discrete and in line with our expectation of being stochastically greater than or equal to the uniform. I can dig up the plot if you’re interested.

110

Sebastian Bischoff @sbischoff.bsky.social · 04/11/2025

Yes, samples from the null across all cells are close to uniformly distributed (leaning towards conservative with more p-values around 1 and fewer around 0).

110

Sebastian Bischoff @sbischoff.bsky.social · 04/11/2025

If you have a suboptimal cell embedding, velotest can actually yield misleading results, as our null distribution is dependent on this embedding and only alternative velocities in the suboptimal embedding are considered. So, one should first validate the cell embedding before applying velotest.

210

Sebastian Bischoff @sbischoff.bsky.social · 04/11/2025

Yes, especially on the COVID dataset, the embeddings don’t seem to work.

110

Sebastian Bischoff @sbischoff.bsky.social · 04/11/2025

Sorry, I misunderstood you here. I would like to test some more “independent” datasets to conclude this, but the performance seems to degrade quite a bit.

110

Sebastian Bischoff @sbischoff.bsky.social · 28/10/2025

If you want to interpret an embedding, first check that the velocities are properly represented. And yes, this is done on the full data. However: velotest's results are specific to the underlying gene expression embedding, and some embeddings might be more suitable than others.

120

Reposted by Sebastian Bischoff

Machine Learning in Science @mackelab.bsky.social · 09/12/2024

Thrilled to announce we have three #NeurIPS2024 papers! Interested in simulating realistic neural data with diffusion models or recurrent neural networks, or in source distribution sorcery? Have a look 👇 1/4

1278

Reposted by Sebastian Bischoff

Jakob Macke @jakhmack.bsky.social · 27/11/2024

Watching the sbi-toolbox grow up, seeing its many uses on a wide range of applications, and experiencing the growth, momentum + team-spirit of the sbi community has been amazing. We now have a short software paper with many new contributions and contributors! So many thanks, and get involved!

0406

Reposted by Sebastian Bischoff

Guy Moss @gmoss13.bsky.social · 10/12/2024

@vetterj.bsky.social and I are excited to present our work at #NeurIPS2024! We present Sourcerer: a maximum-entropy, sample-based solution to source distribution estimation. Paper: openreview.net/forum?id=0cg... Code: github.com/mackelab/sou... (1/8)

openreview.net

Sourcerer: Sample-based Maximum Entropy Source Distribution Estimation

Scientific modeling applications often require estimating a distribution of parameters consistent with a dataset of observations - an inference task also known as source distribution estimation....

13010