🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Is motor preparation an optimal control strategy? If you’re interested in finding out, come check out our work with Calvin Kao and Guillaume Hennequin at #BernsteinConference tomorrow (poster 245)!

We just released the 'mgplvm' python package implementing our manifold Gaussian Process Latent Variable Model (recently accepted at #NeurIPS2020!). Check it out if you’re keen to see if the brain really does live on a non-Euclidean manifold! github.com/tachukao/mgplv…

Our Nature Neuroscience paper is out, first-authored by my student Omid Sani. It develops PSID, a novel dynamic method to dissociate & model behaviorally relevant neural dynamics. Thanks to Salar Abbaspour, Y Wong & collaborator Pesaran Lab. Summary & code below: 1/n nature.com/articles/s4159…

We expanded our manifold GPLVM for latent variable modeling in neuroscience with linear/non-linear tuning, Gaussian/discrete noise, and smooth/iid priors – on both Euclidean and non-Euclidean manifolds! Come check it out at #cosyne21 1-086 (w/ Calvin Kao & Guillaume Hennequin )

Come for Bayesian GPFA which automatically learns the latent dimensionality of neural data; stay for analyses of multi-region & preparatory activity in 30 minute long recordings from a self-paced reaching task! (with Calvin Kao,Jasmine Stone,Guillaume Hennequin) biorxiv.org/content/10.110…

Natural continual learning: success is a journey, not (just) a destination! (new work by Calvin Kao and I with Alberto Bernacchia and Guillaume Hennequin; arxiv.org/abs/2106.08085)

I've surfaced these three Neuromotor Interfaces *internship* job postings. Come work with us at FRL CTRL! (Currently has, or obtaining a Ph.D.) Experimental Systems facebook.com/careers/v2/job… Computational Modeling facebook.com/careers/v2/job… Signal Processing facebook.com/careers/v2/job…

For argmin differentiation of strongly convex problems, implicit diff theoretically achieves better Jacobian estimates than autodiff of unrolled gradient descent iterations as a function of iteration error. Empirically validated on ridge regression below. arxiv.org/abs/2105.15183

Is chaos actually hard to predict? For NeurIPS this year I made a database of 131 known strange attractors, and trained state-of-the-art forecasting models on each one, to try to figure this out (1/N): Paper: arxiv.org/abs/2110.05266 Dataset + Code: github.com/williamgilpin/…

Cornell CS3110 book gets better and better: cs3110.github.io/textbook/cover…. The book now has embedded videos and live coding within the book. Building this content & tooling takes an incredible amount of work. Thank you!

“You never forget how to ride a bike” – but how is that even possible when neural circuits in your brain are in constant flux? Check out our new work on neural stability in the motor system w/ NaamaKadmonHarpaz, Ashesh Dhawale, Steffen Wolff & @BOlveczky! tinyurl.com/zhuz6yhk

Excited to launch the new Allen Institute for Neural Dynamics (tinyurl.com/594b3hme). We are a multidisciplinary group of scientists & engineers organized to understand how brain-wide neural circuits perform the computations underlying flexible behavior.

Following up the announcements from #COP26 that we aim to end deforestation by 2030, I’m co-founding the Cambridge Centre for Carbon Credits to help make this reality by building a trusted marketplace to verify and fund nature-based solutions cam.ac.uk/research/news/…

Why do we see both sequential and persistent activity (ie dynamic coding) in PFC during working memory? In our new paper, we show that the optimal way to load information into working memory results in exactly these observed dynamics. biorxiv.org/content/10.110… 1/5

Cambridge-Columbia collaborative postdoc position available between the labs of Guillaume Hennequin, Máté Lengyel & Daniel Wolpert to study neural network mechanisms underlying the context-dependent (continual) learning of motor repertoires. tinyurl.com/ycxhw6wa. Come work with us!

Ella Batty Decided to verify that this works and wrote a quick jupyter notebook; here it is in case anyone is interested (including a comparison with a simple linear model predicting cos/sin theta): colab.research.google.com/drive/1xlsv7gb…)

Long-term stable motor behaviors are generated by single neuron activity patterns in motor cortex and striatum that are themselves highly stable New from Kristopher Torp Jensen NaamaKadmonHarpaz Ashesh Dhawale Steffen Wolff @BOlveczky Harvard Organismic & Evolutionary Biology HarvardBrainScience nature.com/articles/s4159…

Thrilled that this is finally out!! Can’t thank my co-authors, the three reviewers, and the entire Ölveczky lab enough for all the help along the way - and of course @StephenEglen for running the MPhil course that the project grew out of!

Very excited to share our recent work w/ Calvin Kao & Guillaume Hennequin! We investigated when & why motor preparation arises in RNN models of motor control. Is preparation optimal for RNNs performing delayed reaching tasks? Spoiler: it is! (1/5) biorxiv.org/content/10.110…

1/7 For the past decade, our team at Meta Reality Labs (previously CTRL-labs) has been dedicated to developing a neuromotor interface. Our goal is to address the Human Computer Interaction challenge of providing effortless, intuitive, and efficient input to computers.