Sign in

Mitchell Ostrow

@neurostrow.bsky.social
266 followers 295 following 52 posts

PhD Student at MIT Brain and Cognitive Sciences studying Computational Neuroscience / ML. Prev Yale Neuro/Stats, Meta Neuromotor Interfaces

PostsRepliesMedia
Reposted by Mitchell Ostrow
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
Very excited to announce our most recent paper! TL;DR we made our prior work on comparing dynamics (DSA) much faster and generalizable to diverse data domains and problem settings (1/)
biorxiv.org
A metric for comparing complex systems by their dynamics
Comparisons are fundamental to science: experiment against model, one organism against another, a system against itself across time. Because many systems, from brains to climate, are characterized by ...
17118
Reposted by Mitchell Ostrow
Will Redman @wtredman.bsky.social · 21/09/2026
DSAmaxxing for comparing dynamical systems! Some really great work from @neurostrow.bsky.social on advancing DSA to be able to shed light on complex, noisy neural dynamics AND using DSA as a regularizer to discover novel RNN solutions Really lucky to have been included and involved in this!
082
Reposted by Mitchell Ostrow
Will Redman @wtredman.bsky.social · 21/09/2026
∂DSA finds lots of RNN solutions to the same integration task: Line attractors ✅ Feedforward chains ✅ Oscillations ✅ Unstable dynamics ✅
021
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
As always, code is available at GitHub: github.com/mitchellostr... Huge thanks to my collaborators on this paper @adamjeisen.bsky.social, @leokoz8.bsky.social, Ila Fiete, @wtredman.bsky.social, and prior ones @annhuang42.bsky.social , @satpreetsingh.bsky.social , @kanakarajanphd.bsky.social ! (10/)
github.com
GitHub - mitchellostrow/DSA: Dynamical Similarity Analysis code accompanying the paper "Beyond Geometry: comparing the temporal structure of computation in neural circuits via dynamical similarity ana...
Dynamical Similarity Analysis code accompanying the paper "Beyond Geometry: comparing the temporal structure of computation in neural circuits via dynamical similarity analysis" - mitchel...
060
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
With the formalism in hand, we can now generalize and extend DSA to various problem and data domains! We’ve also sped up the algorithm substantially — in this preprint, we performed 2 million comparisons! (9/)
110
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
We also built a fast, differentiable form of DSA — ∂DSA. With this, we showed we can train RNNs online with DSA to learn novel, never-before-hypothesized solutions to computational tasks! (Stay tuned for future work on this too!) (8/)
171
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
DSA also enables cross-task comparison within the same population via conditionalizing on input or other variables. Using this approach we showed that two different reaching tasks in macaque M1 diverge over years from a shared solution space, and that M1 decouples from PMd at movement onset! (7/)
130
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
DSA identifies for the first time that rodent head direction circuit dynamics are indistinguishable from ring attractor models, and that it distinguishes different stages of sleep in this circuit at different timescales, which is due to the on- and off-manifold dynamics of the circuit! (6/)
120
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
We demonstrate the DSA family's robustness and flexibility on three diverse settings of interest to neuroscience: rodent head direction dynamics, macaque motor cortex, and the application of DSA online during training (5/)
120
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
Here we take that to the extreme and show that DSA is actually a family of methods! Different problems require different notions of similarity and have data formats (spikes, LFP, behavior) or different dynamical structure (input drive, noise, trial structure, natural behavior, etc).(4/)
110
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
DSA, by design, is a modular system--identify a linear representation of your dynamics, then compare with matrix similarity metrics. (3/)
110
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
We previously introduced two related algorithms: Dynamical Similarity Analysis (DSA) and InputDSA, which are data-driven similarity metrics for dynamical systems. openreview.net/forum?id=2Fr... proceedings.neurips.cc/paper_files/... (2/)
openreview.net
Verifying your browser | OpenReview
110
Mitchell Ostrow @neurostrow.bsky.social · 21/09/2026
Very excited to announce our most recent paper! TL;DR we made our prior work on comparing dynamics (DSA) much faster and generalizable to diverse data domains and problem settings (1/)
biorxiv.org
A metric for comparing complex systems by their dynamics
Comparisons are fundamental to science: experiment against model, one organism against another, a system against itself across time. Because many systems, from brains to climate, are characterized by ...
17118
Reposted by Mitchell Ostrow
Kempner Institute at Harvard University @kempnerinstitute.bsky.social · 09/03/2026
NEW from the #KempnerInstitute: InputDSA, a tool to separate intrinsic dynamics from input-driven effects—enabling accurate, efficient comparisons of complex systems with external inputs. Read the #DeeperLearning blog post by @annhuang42.bsky.social & @kanakarajanphd.bsky.social: bit.ly/4bkrvy9
bit.ly
InputDSA: Demixing then comparing recurrent and externally driven dynamics in complex systems - Kempner Institute
We explored how to measure the similarity between two complex systems when they are driven by external inputs, like biological neural circuits or reinforcement learning agents. Our novel method, calle...
0219
Reposted by Mitchell Ostrow
David G. Clark @david-g-clark.bsky.social · 09/03/2026
Very clear & interesting blog post from @annhuang42.bsky.social and @kanakarajanphd.bsky.social on their paper!
081
Reposted by Mitchell Ostrow
Matt Perich @mattperich.bsky.social · 10/03/2026
To assess conserved neural computations, we tested the similarity of neural dynamics with Dynamical Similarity Analysis, an awesome technique for comparing the “rules” of two dynamical systems (arxiv.org/abs/2306.10168). Low dynamical distance, here, implies similar computations. 6/18
1113
Mitchell Ostrow @neurostrow.bsky.social · 23/04/2026
Go talk to Ann about our method! I won’t be there :(
030
Reposted by Mitchell Ostrow
Kanaka Rajan @kanakarajanphd.bsky.social · 11/03/2026
📍 #Cosyne26 bound? Thrilled to have PhD students @annhuang42.bsky.social and Sonja Johnson-Yu representing the Rajan Lab this year. Shout out to co-first authors @satpreetsingh.bsky.social and @neurostrow.bsky.social. Catch their sessions to learn about our newest work 🧠🤖 #NeuroAI #CompSci
Rajan Lab at Cosyne
Thurs., March 12 20:30-23:30 Pavilion 1 Poster Session 1
#1-059: Active electrosensing and communication in MARL-trained weakly electric fish foraging agents
Sonja Johnson-Yu*, Satpreet Singh*, Zhouyang Lu, Aaron Walsman, Federico Pedraja, Denis Turcu, Pratyusha Sharma, Naomi Saphra, Nathaniel Sawtell, Kanaka Rajan
*co-first authors
Sat., March 14 11:45-12:00 Auditorium I Session 7: Dynamical Inference
InputDSA: Demixing then comparing recurrent and externally driven dynamics
Ann Huang*, Mitchell Ostrow*, Satpreet Singh, Leo Kozachkov, Ila Fiete, Kanaka Rajan
*co-first authors
0265
Mitchell Ostrow @neurostrow.bsky.social · 26/01/2026
Accepted to ICLR! see you in 🇧🇷
0131
Reposted by Mitchell Ostrow
Arman Behrad @armanbehrad.bsky.social · 08/01/2026
The original dynamic similarity analysis (DSA) developed by @neurostrow.bsky.social and Ila Fiete is a powerful method to compare trajectories of (nonlinear) neural dynamics between different datasets and models: arxiv.org/abs/2306.10168
172
Reposted by Mitchell Ostrow
Arman Behrad @armanbehrad.bsky.social · 08/01/2026
Wanna compare dynamics across neural data, RNNs, or dynamical systems? We got a fast and furious method🏎️ The 1st preprint of my PhD 🥳 fast dynamical similarity analysis (fastDSA): 📜: arxiv.org/abs/2511.22828 💻: github.com/CMC-lab/fast... I’ll be @cosynemeeting.bsky.social - happy to chat 😉
111834
Mitchell Ostrow @neurostrow.bsky.social · 06/12/2025
Causal to what? we know from biophysics how spikes causally trigger neurotransmitter release, and how neurotransmitters cause PSPs, which trigger spiking in post synaptic neurons etc…
040
Mitchell Ostrow @neurostrow.bsky.social · 04/12/2025
Woah huge!! Congrats
010
Reposted by Mitchell Ostrow
Victor Geadah @vgeadah.bsky.social · 03/12/2025
At #NeurIPS2025! 🎉 Excited to present Conditionally Linear Dynamical Systems (CLDS). We leverage the dependence of neural dynamics on task covariates to yield an interpretable, flexible model of dynamics. Come meet and check it out! 📍: Poster #2209, Hall C,D,E on Thu Dec 4, 11 am–2 pm, PST. 🧵/6
1195
Mitchell Ostrow @neurostrow.bsky.social · 30/11/2025
similarly: arxiv.org/abs/2510.26418
arxiv.org
Chain-of-Thought Hijacking
Large reasoning models (LRMs) achieve higher task performance with more inference-time computation, and prior works suggest this scaled reasoning may also strengthen safety by improving refusal. Yet w...
020
Reposted by Mitchell Ostrow
Adam J. Eisen @adamjeisen.bsky.social · 26/11/2025
13/ 😀Feel free to reach out to discuss this work, or the application of it to your field of study. Or come swing by our poster at #NeurIPS2025. We’d love to chat! 📄 Paper: openreview.net/forum?id=I82... 💾 Code: github.com/adamjeisen/J... 📍 Poster: Thu 4 Dec 11am - 2pm PST (#2111)
openreview.net
Characterizing control between interacting subsystems with deep...
Biological function arises through the dynamical interactions of multiple subsystems, including those between brain areas, within gene regulatory networks, and more. A common approach to...
2113
Mitchell Ostrow @neurostrow.bsky.social · 28/11/2025
Really proud of this project with @adamjeisen.bsky.social - Jacobian estimation is a challenging and generic problem in dynamics, and I’m excited for all the future use cases of our method! See you at NeurIPS 🧠💻
081
Reposted by Mitchell Ostrow
Adam J. Eisen @adamjeisen.bsky.social · 26/11/2025
How do brain areas control each other? 🧠🎛️ ✨In our NeurIPS 2025 Spotlight paper, we introduce a data-driven framework to answer this question using deep learning, nonlinear control, and differential geometry.🧵⬇️
18930
Mitchell Ostrow @neurostrow.bsky.social · 26/11/2025
Also, from a dynamics perspective, directions with very little variance (in a statistical perspective) can still have an outsized effect on the activity on directions with larger variance!
031
Mitchell Ostrow @neurostrow.bsky.social · 19/11/2025
Controversial take: our ICLR reviews actually helped make our paper better
271
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Thanks again to all my amazing collaborators, especially my co-first author @annhuang42.bsky.social !
030
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Public code is here github.com/mitchellostr... , and it is soon to be merged into the DSA package (pip install dsa-metric)
github.com
GitHub - mitchellostrow/DSA at inputdsa
Dynamical Similarity Analysis code accompanying the paper "Beyond Geometry: comparing the temporal structure of computation in neural circuits via dynamical similarity analysis" - GitHub ...
130
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Second, we develop a new similarity metric based in control theory and shape metrics, which is extremely fast and robust (no figure here)! The metric is based on controllability, which measures how easily inputs can arbitrarily move the state of a dynamical system.
130
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
First, we apply subspace id methods from classical control theory to learn input-controlled linear dynamical systems (key in partially observed settings). This is new for the Dynamic Mode Decomposition (DMD) literature, and the method robust to extreme partial observation (12/)
140
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Now for the 🤓 : InputDSA leverages 2 new technical developments (11/)
120
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
We think that inputDSA could be especially useful when experimentalists can perturb a system (e.g with optogenetics) for system identification. (10/)
130
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
As with DSA, inputDSA complements other comparison metrics (@itsneuronal.bsky.social , @mschrimpf.bsky.social ). One important result we found is that even for input-driven dynamics, the original DSA still gives good comparisons, but inputDSA can sharpen them! (9/)
130
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
On two datasets, we apply random perturbations (noise, functions) to the true input, or utilize other task variables when performing inputDSA. We measure the correlation between the surrogate and true scores, finding that in general, inputDSA is quite robust! (8) (shoutout @oliviercodol.bsky.social)
150
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
One more analysis with greater implications: In most neuroscience settings, we don’t know the true inputs to a brain region. When we build models, we apply proxy inputs that we think are related to the true input. With InputDSA, we can evaluate this! (7/) (as in e.g line attractors in hypothalamus))
140
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Second: on @thomas-zhihao-luo.bsky.social recently showed that rat cortical dynamics transition from primarily input-driven to autonomous during a 2-alternative forced choice task. InputDSA corroborates this, showing that cortex becomes less input-controllable across time! (6/)
150
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
On @satpreetsingh.bsky.social ’s Deep RL fly navigation task (from @bingbrunton.bsky.social ’s lab) we show that successful models become more similar to each other across training, while unsuccessful ones diverge in inputDSA score —an Anna Karenina/universality result! (5/)
130
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Let’s look at some cool applications first! We made a lot of technical developments, but I'll save those till the end 🤓 :
120
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
The basic idea of DSA: approximate your dynamics so that comparison is tractable. This is backed by Koopman Operator Theory and relates to work done by @wtredman.bsky.social and Igor Mezic. InputDSA naturally extends DSA—we can compare intrinsic dynamics, the effect of input, or both jointly! (3/)
140
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
We introduce InputDSA, a method that builds on our prior work, Dynamical Similarity Analysis (DSA) to quantitatively compare input-drive dynamical systems! Especially relevant for neuroscience, but it can be applied to any type of time series data ! 🧠 💻 🌴 💨 💵 🔥 (2/)
140
Mitchell Ostrow @neurostrow.bsky.social · 10/11/2025
Our next paper on comparing dynamical systems (with special interest to artificial and biological neural networks) is out!! Joint work with @annhuang42.bsky.social , as well as @satpreetsingh.bsky.social , @leokoz8.bsky.social , Ila Fiete, and @kanakarajanphd.bsky.social : arxiv.org/pdf/2510.25943
47124
Reposted by Mitchell Ostrow
Sam Cheyette @samcheyette.bsky.social · 14/10/2025
Very excited to share a new preprint that’s been brewing for a long time! This work was led by the exceptional @traceym.bsky.social, and made possible by a developmental + comparative + computational dream team. osf.io/preprints/ps...
osf.io
OSF
1155
Mitchell Ostrow @neurostrow.bsky.social · 09/07/2025
This doesn't say anything about how the attractors is instantiated, ie the equation itself (let alone its mapping to the biology, which is another criterion needed for a mechanism according to Craver). I'm fine with this claim if it's what the post means!
130
Mitchell Ostrow @neurostrow.bsky.social · 09/07/2025
Perhaps what is meant by 'attractors aren't mechanisms' is that you can write down a large number of equations that are attractors (e.g. any diffeomorphism phi that transforms the system dxdt = -x while preserving its asymptotic behavior, also known as a conjugacy).
110
Mitchell Ostrow @neurostrow.bsky.social · 08/07/2025
I should clarify—in my example the system is the linear dynamical equation. When you say system, what are you referring to?
100
Mitchell Ostrow @neurostrow.bsky.social · 08/07/2025
I guess I’m still not following, especially wrt usage of the words caused, composition and organization. Maybe we can use an example? Take dxdt=-x. the dynamics in this system is the decay to the attractor. So I’m not clear how to distinguish the two…
100