Ling-Qi Zhang

Bayesians in shambles

📢 Students & postdocs in #TheoreticalNeuroscience 👉 Join us this November for a "by the students, for the students" workshop featuring whiteboard-style tutorials, in-depth discussions, and research talks.

✈️ Travel, lodging & registration covered

Apply by Aug. 5 @ janelia.news/THE25

What is the probability of an image? What do the highest and lowest probability images look like? Do natural images lie on a low-dimensional manifold?
In a new preprint with Zahra Kadkhodaie and @eerosim.bsky.social, we develop a novel energy-based model in order to answer these questions: 🧵

Thanks!! There are no major changes in the main text re results. We did refine the vignetting analysis, but it didn't make a huge difference in the end results. We also added a polar ROI analysis in the supplement, complementary to the eccentricity ROI in Fig. 5.

Simons Foundation Launches Collaboration on Ecological Neuroscience Simons Foundation Launches Collaboration on Ecological Neuroscience on Simons Foundation

Very excited to be part of this. Huge thank you to the @simonsfoundation.org for their support and vision.

www.simonsfoundation.org/2025/04/24/s...

Also, job openings in this context! Please apply and forward to anyone who may be interested!
hr.myu.umn.edu/jobs/ext/367...

Thanks Megan!!

Same here!!

Space and time in visual context - Nature Reviews Neuroscience Sensory stimuli are interpreted in light of their spatial and temporal context, which can sometimes lead to visual illusions. Schwartz, Hsu and Dayan consider the possible functional benefits of conte...

And I'm very excited to see how this theoretical framework (the original credit goes to www.nature.com/articles/nrn...) might explain many other illusions (e.g., motion aftereffect) in a principled way! (6/6)

Adaptation optimizes sensory encoding for future stimuli Author summary Prolonged exposure to a fixed stimulus causes sensory neurons to adapt. In this study, we uncover some of the functional benefits of adaptation for the visual system. We first quantifie...

Lastly, our results, together with some recent work such as journals.plos.org/ploscompbiol... and www.biorxiv.org/content/10.1... @mariodipoppa.bsky.social, are painting a compelling picture of how contextual effects (both spatial and temporal) can be interpreted as adaptive efficient coding. (5/6)

In addition, the change in encoding is sufficient to predict the behavioral characteristics of the tilt illusion using a Bayesian observer model. (4/6)

With an oriented surround, encoding precision is locally increased for stimuli similar to the surround orientation, specifically in higher areas of the visual cortex. This encoding change also reflects the surround-conditioned orientation statistics in natural scene. (3/6)

We measured sensory encoding precision (in terms of Fisher information) from both behavioral and neural data in an orientation estimation task. At baseline, encoding reflects the natural scene statistics of orientation. (We observe the cardinal bias in different parts of the visual cortex!). (2/6)

The tilt illusion arises from an efficient reallocation of neural coding resources at the contextual boundary | PNAS The tilt illusion—a bias in the perceived orientation of a center stimulus induced by an oriented surround—illustrates how context shapes visual pe...

Very happy to see our work finally in print!
www.pnas.org/doi/10.1073/...

TLDR: Tilt illusion is not a bug, but a feature of a well-designed visual system that maximizes information capacity adaptively based on spatial context. (1/6)

many many tiny contextual boundaries!

And 2) If we can create a surround that REALLY modulates the representation at the boundary, we might be able to break the segmentation cue, and see an illusion as you suggested. This will require an image synthesis method for surround modulation, but people are working on this type of idea!

This does lead to some "really cool if true" predictions: 1) If we gradually increase the grey center aperture, the illusion strength should first increase (as we are removing orientation information not modulated by the surround), and then decrease (once we are in the "center-surround" boundary).

Great question! My explanation would be that there are strong enough segmentation cues in the stimulus for the visual system to integrate the orientation information across the center, to form a holistic orientation estimate.

AI@MBL: Machine Learning for Microscopy Image Analysis | Marine Biological Laboratory The goal of this course is to familiarize researchers in the life sciences with state-of-the-art deep learning techniques for microscopy image analysis and to introduce them to tools and frameworks th...

How can AI help you to analyze your microscopy images?

Find out at the "Machine Learning for Microscopy Image Analysis" course at the @mblscience.bsky.social in Woods Hole from Aug 22-Sep 6!

Applications are due 🚨 April 30 🚨

Read on for more! 👇

www.mbl.edu/education/ad...

To stop intimidation, Eisgruber needs your help When targeting one institution is a bald attempt to divide and conquer, as is the case with Columbia, it is essential for a coalition or group of institutions and individuals to stand up together.&nbs...

Here at Princeton, we’re amplifying our university president’s leadership. It’s time for everyone - faculty, students, staff - to speak out.

tl;dr: rights can’t be blackmailed.

From me and science colleagues:
www.dailyprincetonian.com/article/2025...

Columbia Concedes to Trump’s Demands After Federal Funds Are Stripped The administration has moved to cut $400 million in federal funding to the university without changes to its policies and rules.

Columbia University has failed to defend academic freedom & freedom of speech & allowed a great institution to be bullied & extorted into submission by an administration hellbent on contorting higher education to their will.

The AAUP condemns this capitulation. We can & must fight back.

Using funding to 'force concessions' threatens institutions, Princeton president says Columbia University agreed to comply with a series of demands from the Trump administration about how it will handle protests, antisemitism and academic departments. The university faced a deadline to...

"And I think once you make concessions once, it's hard not to make them again." Princeton President Chris Eisgruber's warning to universities applies to law firms, companies, foundations, nonprofits, and more. www.pbs.org/newshour/sho...

Pro-tip for university administrators: when the government illegally cancels your grants, your first response shouldn't be "you kind of have a point."

old Soviet joke for our times:

Guy stops by the newsstand every day, scans the front page, doesn’t buy the paper. One day the vendor asks what he’s up to.

Guy says: “looking for an obituary.”

Vendor says “those are towards the back of the paper, comrade.”

Guy says: “not the one I’m looking for.”

I got laid off today, with the rest of 18F.

18F was an elite federal software shop. We made gov't websites work better, more efficiently for the American people. We saved taxpayers from getting screwed over by contractors. And were fired for it.

We made this website to tell our story:
18f.org

Portuguese editorial cartoonist Zez Vaz reaches back to Tiananmen Square to call on American defiance.

I maxed out my credit card, emptied my savings account, and took out a loan to move from Alabama to Bethesda, MD. I don’t even qualify for unemployment since I’ve only worked at NIH for a month. I will be financially and medically devastated.

Seeking suggestions for anywhere that’s hiring!!

Large Concept Models: Language Modeling in a Sentence Representation Space – Janelia CVML Summary of and thoughts on Meta’s recent paper

We read "Large Concept Models" in our reading group. It models language at the sentence resolution - "concepts". We had lots of thoughts about it! You can read them in our new Janelia CVML blog post:
janelia-cvml.github.io/blog/posts/L...
#ML #AI #Janelia @adjavon.bsky.social @bogovicj.bsky.social

This comment on the NYT site sounds spot-on, and it is why I disagree with my colleagues who say: "We'll know more in a few weeks, let's wait and see how things shake down before we take any action.":
"As someone who grew up under an authoritarian regime, I can offer one crucial piece of advice: ...

"The first rule of research administration is [generally scientists don't know much about it]" Time for that to change

Dear US scientists,

We've been hearing a lot of questions of the form
"Why aren't universities doing anything!? They need to speak out!"

We can explain.

It's bad news. We're on our own. But understanding why can help us all figure out our plan.

Read along in thread 1/x 🧪