Harsha Gurnani

New preprint from the lab! Ábel Ságodi developed a theory of approximating dynamical systems that goes beyond finite time. #theoreticalNeuroscience
Follow @neurabel.bsky.social

Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees. . arxiv.org/abs/2602.08640

Super cool!
Was wondering if this work by @chaichinta.bsky.social would be relevant to think about the GS spiking?
www.pnas.org/doi/abs/10.1...

Thrilled to finally share this work! 🧠🔊

Using a new reinforcement-free task we show mice (like humans) extract abstract structure from sound (unsupervised) & dCA1 is causally required by building factorised, orthogonal subspaces of abstract rules.

Led by Dammy Onih!
www.biorxiv.org/content/10.6...

(11/11) In this new normal, costs are individually adjusted to a consumer’s maximum threshold and wages to a worker’s minimum floor.
The next time you see a price, know that it may not reflect what the item is worth—but what the algorithm believes *you* are worth.

(1/11) If you live in NY, you’ve probably started seeing a new warning: “THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA.” This mandatory disclosure went into effect late last year, and it’s the first attempt by a US state to grapple with a new generation of surveillance pricing.

Oo looks super relevant for something I'm puzzling over myself, will definitely give this a read. Congrats!

New paper out at PNAS: www.pnas.org/doi/10.1073/...
Revisiting the high-dimensional geometry of population responses in the visual cortex with @jpillowtime.bsky.social. The review took forever because a reviewer was doubtful our new estimator can infer eigenvalues beyond the rank of the data! (1/6)

I truly enjoyed writing this profile. Eva's personal trajectory from the clinic to academia highlights the ongoing importance of《fundamental science》for tackling mental health and addiction. Science enables the shift from "labels" to "mechanisms" and leads to more tailored, powerful approaches.

Thank you for the great summary, @thetransmitter.bsky.social, and the shout-outs from scientists we admire. There is so much we don’t know about how sex hormones modulate behavior and we’ll keep exploring!

Another nail in the coffin for PCA?

- doesn’t linearize, distorting similarity metrics
- is biased by temporal jitter across epochs
- may miss important dimensions for transient amplification

If you think there is a state space, use a state space model!

Feedback control of recurrent circuits imposes dynamical constraints on learning Neural activity has been observed to lie on low-dimensional manifolds, constraining what behaviors are easier or harder to learn. We propose that beyond this geometric constraint, learning on fast tim...

Looking forward to reading the paper! Struggled with the same mismatch for a long time. In this paper, we showed that looking at underlying dynamical subspaces is going to give us different answers than reducing dim with PCA, eg. for understanding learning variability
www.biorxiv.org/content/10.1...

Note: this has a computational scientist track! It’s for you, too, math nerds!

Different learning algorithms achieve shared optimal outcomes in humans, rats, and mice Animals must exploit environmental regularities to make adaptive decisions, yet the learning algorithms that enabels this flexibility remain unclear. A central question across neuroscience, cognitive science, and machine learning, is whether learning relies on generative or discriminative strategies. Generative learners build internal models the sensory world itself, capturing its statistical structure; discriminative learners map stimuli directly onto choices, ignoring input statistics. These strategies rely on fundamentally different internal representations and entail distinct computational trade-offs: generative learning supports flexible generalisation and transfer, whereas discriminative learning is efficient but task-specific. We compared humans, rats, and mice performing the same auditory categorisation task, where category boundaries and rewards were fixed but sensory statistics varied. All species adapted their behaviour near-optimally, consistent with a normative observer constrained by sensory and decision noise. Yet their underlying algorithms diverged: humans predominantly relied on generative representations, mice on discriminative boundary-tracking, and rats spanned both regimes. Crucially, end-point performance concealed these differences, only learning trajectories and trial-to-trial updates revealed the divergence. These results show that similar near-optimal behaviour can mask fundamentally different internal representations, establishing a comparative framework for uncovering the hidden strategies that support statistical learning. ### Competing Interest Statement The authors have declared no competing interest. Wellcome Trust, https://ror.org/029chgv08, 219880/Z/19/Z, 225438/Z/22/Z, 219627/Z/19/Z Gatsby Charitable Foundation, GAT3755 UK Research and Innovation, https://ror.org/001aqnf71, EP/Z000599/1

paper🚨
When we learn a category, do we learn the structure of the world, or just where to draw the line? In a cross-species study, we show that humans, rats & mice adapt optimally to changing sensory statistics, yet rely on fundamentally different learning algorithms.
www.biorxiv.org/content/10.1...

Who wore it better? Our new paper shows that rat OFC supports Bayesian inference of hidden states! With neural correlates of inferred state transitions at the level of single neurons and population-level latent factors. www.sciencedirect.com/science/arti...

Estrogen modulates reward prediction errors and reinforcement learning - Nature Neuroscience Dopamine encoding of reward prediction errors naturally fluctuates over females’ reproductive cycles with estrogenic signaling due to reduced expression of dopamine reuptake proteins.

Thrilled to share our new paper, out now in @natneuro.nature.com, uncovering how estradiol, the most potent estrogen, modulates reinforcement learning and reward prediction errors across biological levels. www.nature.com/articles/s41...

#blueprint 1/7

a man wearing a white shirt and tie smiles in front of a window ALT: a man wearing a white shirt and tie smiles in front of a window

I've been waiting some years to make this joke and now it’s real:

I conned somebody into giving me a faculty job!

I’m starting as a W1 Tenure-Track Professor at Goethe University Frankfurt in a week (lol), in the Faculty of CS and Math

and I'm recruiting PhD students 🤗

Love this!

Amazing news, congratulations!!

*10y ago oops

I love watching mentees transform wonderment into knowledge & growth, it reminds me of our collective responsibility & power 👩🏻‍🏫 Also surreal that it was 10y that my UG experience was transformed by amazing, kind mentors like Eve, Rishi and Tim 😊 @timothyoleary.bsky.social
@cnl-mbu-iisc.bsky.social

Qualifications:
PhD in CompNeuro/ML/related field.
Strong background in ML and programming
Previous research experience in RL, probabilistic ML, or statistical methods for neural data analysis.
Strong motivation & ability to work independently and collaboratively in a multidisciplinary environment.

Postdoctoral Associate (Savin Lab) - HigherEdJobs Jobs in higher education. Faculty and administrative positions at colleges and universities. Updated daily. Free to job seekers.

Application:
To apply please email csavin@nyu.edu:
- your CV
- contact details of at least 2 references
- a brief (one page maximum) description of why are you interested in the position and how your expertise fits the call.

Or Job Ad here: www.higheredjobs.com/faculty/deta...

#neuroskyence

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

Job post alert!👩‍🔬 Postdoctoral fellow in CompNeuro/ML in the Artificial and Biological Computation lab at NYU (csavin.wixsite.com/savinlab).
Exciting opening in the Data and Theory team in the new Simons Collaboration in Ecological Neuroscience!
www.simonsfoundation.org/2025/04/24/s...
#neuroskyence

🤣😭

Motor cortex flexibly deploys a high-dimensional repertoire of subskills Skilled movement often requires flexibly combining multiple subskills, each requiring dedicated control strategies and underlying computations. How the motor system achieves such versatility remains u...

Of potential interest to those keen on motor control and/or multi-task networks. Congrats to Elom and Eric.
www.biorxiv.org/content/10.1...

as an aside, arealization does not need to imply modules in a feedforward chain; in fact these studies very much reconcile recurrent computation with dissociable contributons.

Integrator dynamics in the cortico-basal ganglia loop underlie flexible motor timing Flexible control of motor timing is crucial for behavior. Before volitional movement begins, the frontal cortex and striatum exhibit ramping spiking activity, with variable ramp slopes anticipating mo...

A recent example from the Inagaki lab: www.biorxiv.org/content/10.1... and a couple of classics from Brody lab: elifesciences.org/articles/34929 and elifesciences.org/articles/05457 use perturbations to show discriminable roles despite similar representations.

Congrats!! Excited to dig into this!

Differential kinematic coding in sensorimotor striatum across behavioral domains reflects different contributions to movement - Nature Neuroscience Hardcastle and Marshall et al. show that striatal function is domain specific, required for task-related but not spontaneously expressed movements. This functional distinction is reflected in starkly ...

Excited to announce that my first postdoc paper is now online!

Links:
www.nature.com/articles/s41...
rdcu.be/eAcN7

In it, we examine the perennial question: what changes in the brain when learning a new motor skill?

Read more below to find out 👇

Excited to share this profile + interview with Dr. Emily Jacobs! Amazing work on human precision brain imaging across hormonal fluctuations, and such an inspiring story! I had such a great time interviewing her!