anirbanray.bsky.social - Profile | ThreadSky | a Reddit-style client for Bluesky

Latent Labs comes out of stealth today with $50M funding. Our goal? To push the frontiers of generative biology, giving partners instant access to tools capable of accelerating drug design

submitted 15 days ago • 3 comments

New startup by DeepMind alums - Latent Labs is building AI foundation models to “make biology programmable,” and it plans to partner with biotech and pharmaceutical companies to generate and optimize proteins techcrunch.com/2025/02/12/f...

submitted 15 days ago • 0 comments

Whoop whoop 🥳

submitted 16 days ago • 0 comments

Great interview with @jascha.sohldickstein.com about diffusion models! This is the first in a series: similar interviews with Yang Song and yours truly will follow soon. (One of these is not like the others -- both of them basically invented the field, and I occasionally write a blog post 🥲)

submitted 17 days ago • 0 comments

📝 FeatureForest: the power of foundation models, the usability of random forests. Nomen est omen! We use features of the biggest models and train a handy random forest based pixel classifier with them. 👍 Hit bioRxiv recently… check it out! www.biorxiv.org/content/10.1...

submitted 17 days ago • 3 comments

Immagine you could image two cellular structures in the same fluorescent channel and still reliably get them separated afterwards… What would you do with this? Now… what would you do if that also worked with 4 structures at once? 👇 #MicroSplit #preview🧵

submitted 20 days ago • 6 comments

Interested in neural operators? Check out this nice blog post that breaks them down by Magdalena Schneider and Michael Innerberger janelia-cvml.github.io/blog/posts/N... @hhmijanelia.bsky.social #deeplearning

submitted 21 days ago • 0 comments

🚨Check it out! 🚨 Submissions are now open for the CellMap Segmentation Challenge! Signup here: cellmapchallenge.janelia.org

submitted 22 days ago • 0 comments

Pretty nice read!

submitted 24 days ago • 0 comments

🚨🚨 MEGA JOB ALERT 🚨🚨 Independent Group Leader Positions in Computational Biology @humantechnopole.bsky.social! Are you ready to start your own lab? Do you know someone who is? Repost this + share with everyone who might want to know about it. Thanks!!! 🙏 More details below... check it out! 🧵 1/3

submitted 30 days ago • 1 comment

Welcome @alex-krull.bsky.social to Bluesky 🎊

submitted 33 days ago • 1 comment

Folks are always surprised to learn that NVIDIA has been working on a foundation model for cell segmentation called Vista 2D that can be trained on a variety of cell imaging outputs, including brightfield, phase-contrast, fluorescence, confocal, or electron microscopy. Blog in thread 🧵

submitted 34 days ago • 1 comment

MedicoSAM is here! 🥳 Our vision foundation model shows promising results for interactive segmentation, for both 2d & 3d medical images, across several tasks and imaging modalities! Don't take my word for this, check us out & try to "segment anything" on medical images! 😉🧵

submitted 36 days ago • 0 comments

📢PSA: #NeurIPS2024 recordings are now publicly available! The workshops always have tons of interesting things on at once, so the FOMO is real😵‍💫 Luckily it's all recorded, so I've been catching up on what I missed. Thread below with some personal highlights🧵

submitted 36 days ago • 1 comment

If you aren’t already using ollama.com, you’re missing out on all the latest LLM models and their crazy performances. Ollama has been the best democratization of the LLM models and helps you to run these models offline with ease. Very impressive!

submitted 37 days ago • 0 comments

Read a cool paper recently 👇 arxiv.org/abs/2501.00744 TL;DR: metrics like FID rely on assumptions of normal distribution which might not be optimal to capture heavy tail behavior of distributions leading to inaccurate evaluations. This provides a solution.

submitted 37 days ago • 0 comments

Congratulations @harschnitz.bsky.social 🎊🎊

submitted 38 days ago • 1 comment

Can you spot the difference between the cells on the left and right? Read on for the solution (and a preprint announcement)… 👇 1/6

submitted 43 days ago • 1 comment

The best microscopy course on the planet is accepting applications!

submitted 48 days ago • 0 comments

Google’s NotebookLM now lets you ask questions as the generated podcast is being played. Like if you’re there as a live audience. If this doesn’t blow your mind, I don’t know what will. Only if Google had a master marketeer like @sama.

submitted 60 days ago • 0 comments

iPhone shortcuts (if you use them well) are truly amazing. From a company dictating how their phones should be used to letting users “pipeline” functions, they’ve come a long way.

submitted 70 days ago • 1 comment

A post by @cloneofsimo on Twitter made me write up some lore about residuals, ResNets, and Transformers. And I couldn't resist sliding in the usual cautionary tale about small/mid-scale != large-scale. Blogpost: lb.eyer.be/s/residuals....

submitted 71 days ago • 3 comments

Schrödinger Bridge Flow for Unpaired Data Translation (by @vdebortoli.bsky.social et al.) It will take me some time to digest this article fully, but it's important to follow the authors' advice and read the appendices, as the examples are helpful and well-illustrated. 📄 arxiv.org/abs/2409.09347

submitted 72 days ago • 0 comments

The slides of my NeurIPS lecture "From Diffusion Models to Schrödinger Bridges - Generative Modeling meets Optimal Transport" can be found here drive.google.com/file/d/1eLa3...

submitted 74 days ago • 9 comments

This paper doesn't get enough credit for what it's worth. This consequential papers is the basis of image quality metrics interpretation for generative models. arxiv.org/abs/1711.06077

submitted 75 days ago • 0 comments

The code for Simplified and Generalized Masked Diffusion for Discrete Data (Jiaxin Shi et al) has been released and a lecture by @arnauddoucet.bsky.social on this topic is also available! 🐍 Code: github.com/google-deepm... 📄 Article: arxiv.org/abs/2406.04329 📼 Video: www.youtube.com/watch?v=qj9B...

submitted 76 days ago • 0 comments

This talk by seems to be the most interesting at #NeurIPS2024. I wish the slides were available already 🥲

submitted 76 days ago • 0 comments

Deepinv v0.2.2 is out! documentation: deepinv.github.io repo: github.com/deepinv/deep... v0.2.2 updates: 📚 New documentation 📊 Metrics classes 🧠 Advanced MRI: 3D and multicoil support 🔥 Bregman potentials 🌀 Advanced transforms: diffeomorphisms, time transforms 🖼️ New datasets ⚡ L12 prior

submitted 78 days ago • 0 comments

Another great resource on Flow Matching from Meta AI github.com/facebookrese...

submitted 80 days ago • 0 comments

After two years, our paper on generative models for structure-based drug design is finally out in @natcomputsci.bsky.social www.nature.com/articles/s43...

submitted 80 days ago • 2 comments

🔉New paper "Generalized Recorrupted-to-Recorrupted" 🔉 with @bemc22.bsky.social and Jorge Bacca - Generalizes R2R self-supervised loss for noise belonging to the exponential family. - Shows asymptotic equivalence with SURE. paper: arxiv.org/abs/2412.04648 code: github.com/bemc22/Gener...

submitted 80 days ago • 1 comment

Unpopular opinion: Social media shorts are ruining attention spans by trapping you in an unconscious spiral of endless scrolling. If you’re going to procrastinate, it’s far better to watch a focused, hour-long video on “How Things Work” than to get lost in mindless shorts. Attention economy sucks.

submitted 82 days ago • 1 comment

Thanks all who helped out and commented in this thread with tips and tricks! I gave a talked recently on some work we have been doing in the lab that incorporated aspects from several of the responses. docs.google.com/presentation... #DiffusionModels #FlowMatching #StochasticInterpolants

submitted 87 days ago • 1 comment

Better VQ-VAEs with this one weird rotation trick! I missed this when it came out, but I love papers like this: a simple change to an already powerful technique, that significantly improves results without introducing complexity or hyperparameters.

submitted 87 days ago • 1 comment

A common question nowadays: Which is better, diffusion or flow matching? 🤔 Our answer: They’re two sides of the same coin. We wrote a blog post to show how diffusion models and Gaussian flow matching are equivalent. That’s great: It means you can use them interchangeably.

submitted 87 days ago • 7 comments

Important to keep in mind when reading about diffusion or flow matching: - diffusion models's Probability Flow ODE is not optimal transport. - Composition of optimal transports is not optimal. - Flow matching with OT conditional paths doesn’t provide optimal transport.

submitted 90 days ago • 2 comments

Can one ever get used of reading such an article about oneself without feeling weird? It’s much worse than hearing your own voice from a recording… 😊 Thanks Helen Zenner for all the work you put into this! 🙏

submitted 92 days ago • 0 comments

VAEs vs GANs will go down in history, much like Bayesian vs Frequentist learning from long gone years. Let us enjoy this moment for what it is — a beautiful moment in time. Too soon it will be gone. Blue skies are smiling on all of us today. Nothing but blue skies do we see.

submitted 92 days ago • 2 comments

New blog post on flow matching: dl.heeere.com/cfm/ Contains some nice visuals too!

submitted 93 days ago • 3 comments

I just realized how cool and consequential this is. This might be the best thing I have seen in 2024 🙌👏

submitted 93 days ago • 0 comments