read image description
ALT
6,213
10.4K
60.2K
Follow
Click to Follow jpillowtime
Jonathan Pillow
@jpillowtime
Computational neuroscientist @ Princeton. Into brains, math, machine learning. Brain is a computer 4eva.
Jonathan Pillow’s Tweets
I am delighted to announce that the "real" camera-ready version of my new book, "Probabilistic Machine Learning: Advanced Topics", is now available. It will appear in print this summer, but it is already freely available online at probml.github.io/book2.
21
486
2,317
Show this thread
Blog post from showing how you can adapt his recently-developed estimator for R^2 to get accurate estimates of representational drift (i.e., that are robust to limited # of trials / response noise)!
Quote Tweet
@jpillowtime LAB BLOG POST! We created an estimator of tuning curve similarity (jneurosci.org/content/42/50/). Found another, RDI, in the awesome paper (nature.com/articles/s4146) by @GoardMichael @tylermarks. Compared in simulation, the magnitude of drift could be much larger! 1/3
Show this thread
11
LAB BLOG POST! We created an estimator of tuning curve similarity (jneurosci.org/content/42/50/). Found another, RDI, in the awesome paper (nature.com/articles/s4146) by . Compared in simulation, the magnitude of drift could be much larger! 1/3
2
13
46
Show this thread
The Christmas jingle we didn't know we needed! 😂
Quote Tweet
6
🚨🚨🚨#TWEEPRINT TIME🚨🚨🚨
and I are thrilled to share our investigation of motor cortical dynamics using optogenetic and electrical perturbations, now on biorxiv.org/content/10.110
Full thread at djoshea.notion.site/Twitter-Thread
9
92
355
Show this thread
Oh wow, it doubles down when asked straight up what 25x75 is! 😂
1
1
10
Show this thread
Maybe we don't need to worry about chatGPT taking our jobs just yet.
3
26
Show this thread
I'm presenting work on "Non-exchangeability in Infinite Switching Linear Dynamical Systems" (w/ ) at the #NeurIPS22 GPSMDMS workshop.📍Friday, Room 387 @ 3:45pm.
Come for parsing processes, state-space models, and a sprinkle of PDEs!
🔗: gp-seminar-series.github.io/neurips-2022/a
1
4
15
Excited to be presenting our work on inverse reinforcement learning for modeling animal decision-making with and at #NeurIPS2022! Come by our poster in Hall J #607 b/w 11-1p to hear more. We will also be doing a virtual deep dive on Dec 7, 1:15 EST.
1
7
44
Show this thread
Excited to present “Dynamic Inverse Reinforcement Learning for Characterizing Animal Behavior” today at #NeurIPS2022. This is joint work with and . Come by poster 607 between 11am and 1pm today and say hi, or catch us next week at the virtual QA!
1
10
47
Show this thread
#NeurIPS2022 papers are publicly available, and I am happy to share ours with and : "Extracting computational mechanisms from neural data using low-rank RNNs" openreview.net/forum?id=M12au
Let's 🧵
1
18
119
Show this thread
Interested in modeling biases in your psychophysical data with a Bayesian ideal observer model? Got a prior that is clearly NOT Gaussian, but you're not willing to give up on your favorite squiggly buddies just yet? 1/3
1
18
50
Show this thread
We're now accepting applications for the 2023 Summer Internship Program! We're funded and we offer a generous stipend, housing, meals, and travel (including to )! Learn more here and apply by 2/1/23! bit.ly/3BSHSQC
2
145
207
Big congrats to my brother Thomas, who PR'd in the Chicago Marathon this morning in 2:36! (5:58 / mile pace). Way to go, bro!!!
3
63
Paper is out! A new method to correct for motion induced artifacts in fluorescence intensity time-series traces of calcium imaging. doi.org/10.1371/journa Congrats to . Fun collaboration w/ !
Quote Tweet
Do you perform optical imaging in behaving animals? An animal's behavior can introduce motion artifacts into fluorescent time traces. Our new method TMAC works by correcting artifacts *after* spatial alignment (1/4)
@jpillowtime @AndrewLeifer @kevinschen13
doi.org/10.1371/journa
Show this thread
read image description
ALT
8
37
Welcome to Assistant Prof ! His lab will study how the brain processes information gathered by the senses, creating computer models to capture the step-by-step calculations across neural circuits.
2
6
59
In Feb 2022, my 6yo had a life threatening incident of anaphylaxis in Mexico. We gave him 2 epi-pens and got him to the ER fast. The ER visit cost us $1,000 in cash.
As soon as we got home, I submitted all required receipts to for reimbursement. However,
17
250
393
Show this thread
The paper also introduces the Temporal Relevance Determination (TRD) prior, a prior for temporal receptive fields or spike-history filters, formed from a Gaussian process prior with logarithmically scaled time axis:
1
1
13
Show this thread
Delighted to have this paper on low-rank receptive field estimation (which we started in 2016) online at last!
Quote Tweet
Excited to share some of my pre-grad school work with @jpillowtime, @RudamntaryNeuro and @GregDField . We develop a Bayesian method for estimating space-time separable receptive fields that scales well to high-d settings and correlated stimuli. Code at github.com/pillowlab/VLR-
1
5
66
Show this thread
Excited to share some of my pre-grad school work with , and . We develop a Bayesian method for estimating space-time separable receptive fields that scales well to high-d settings and correlated stimuli. Code at github.com/pillowlab/VLR-
Quote Tweet
Scalable variational inference for low-rank spatio-temporal receptive fields biorxiv.org/cgi/content/sh #biorxiv_neursci
4
26
129
New work (led by ) introduces "knockout training", a new method for training a DNN in 1-1 correspondence with neurons using (perturbed) behavioral data only!
Quote Tweet
Excited to share a new manuscript!
Deep nets are great at predicting visual neurons. Yet, they are unable to tell us which artificial neuron directly corresponds to a biological neuron… until now!
biorxiv.org/content/10.110
(yes, that is indeed a fictive female fly, good guess!)
Show this thread
0:09
22.8K views
2
29
Quote Tweet
Update of potential interest: I'll be starting a comp neuro group at Cold Spring Harbor Laboratory @CSHL in Sept 2022!
Our group will build closed-loop machine learning techniques and interpretable, data-driven models to understand the step-by-step computations of the brain.
Show this thread
1
19
The challenge is fully functional. Let's predict some neurons! Please retweet!
Quote Tweet
We have just launched our SENSORIUM competition to find the best model of mouse primary visual cortex! 
Have a look at out white paper on #arxiv where we describe the data and the competition tracks in detail: arxiv.org/abs/2206.08666.
#NeurIPS2022 Competition track.
Have a look at out white paper on #arxiv where we describe the data and the competition tracks in detail: arxiv.org/abs/2206.08666.
#NeurIPS2022 Competition track.Show this thread
1
15
18
Show this thread
Dear trainees, two emails is not stalking. If you write to a PI requesting something and they don’t get back to you within a reasonable time frame (say 1-2 weeks) please don’t hesitate to contact them again. They may have meant to but forgot. 1/
18
284
1,684
Show this thread
anyone else have university email through microsoft outlook that they forward to a gmail account that all of the sudden is leading to tons of stuff in your spam folder as 'undeliverable' (but actually delivered...just to spam?) [have asked IT locally too]
7
2
9
Show this thread
Clearly need more ads for track & field!
Quote Tweet
Where dreams come true 
50 days to go until the #WorldAthleticsChamps!
50 days to go until the #WorldAthleticsChamps!
2:11
46.3K views
2
New at SCGB news: New efforts explore how internal states influence decision-making, featuring research from , and
10
18
The 90s really were amazing.
23
45
316
Best video game sound track ever???
Quote Tweet
youtube.com/watch?v=b4oHa3
Show this thread
1
1
3
My new preprint with ! An animal's movements during optical neural recordings can induce motion artifacts in the data. We introduce TMAC, a method that uses an activity-independent channel to remove these artifacts.
19
57
Preprint time! Exciting work on manifestations of system-level pressures on dynamic coding mechanisms at the single-neuron level. Great summary in this thread, and for a visual summary ../2
Quote Tweet
Long-time coming preprint led by @vgeadah. We explore why single neurons in the brain may have such diverse and adaptive input-output properties, and how these can emerge from system-level task requirements. What we find is pretty exciting…tinyurl.com/yys4dedb
Show this thread
2
1
17
Show this thread
Check out the power of team science to take on reproducibility!
Quote Tweet
Pre-print announcement! We are excited to share our new pre-print addressing an important issue in systems neuroscience: reproducibility. 
(1/8)
biorxiv.org/content/10.110Show this thread
1
4
New work with Lyndon Duong and Eero Simoncelli on unifying perceptual measurements with Fisher information.
Quote Tweet
A common framework for discriminability and perceived intensity of sensory stimuli biorxiv.org/cgi/content/sh #biorxiv_neursci
2
13
53
Show this thread
Excited to share that I will be joining the Dept of BME at BU!
Lab will focus on models of neural circuits in search of theoretical principles of computation and machine learning methods for analyzing neural population dynamics!
Keep an eye out for exciting news to come!
Quote Tweet
We are delighted to announce that @briandepasquale will be joining @BostonU_BME in January. Brian is a superb addition to @buCSNneuro and @NeurosciBU. Welcome, Brian!
26
18
200
Did you know that our brains, pupils, gaze and heart rate synchronize with other people when we watch videos? But why? And which signals should or should not synchronize?
We test a new theory on 150+ people, with @lucas_c_parra. Out in : doi.org/10.1093/pnasne
1/13
10
121
474
Show this thread
Join us for a virtual talk by the incredible ✨Dr. Wendy Suzuki ()✨ as she walks us through the formative parts of her science journey!
🌟 Mon. April 18th at 4pm ET
🌟 Open to everyone
Register here: bit.ly/3iMMmyZ
read image description
ALT
10
24
New paper with analyzing how running affects neural responses in visual areas beyond V1. We found that running increases the reliability of stimulus encoding (even in neurons and areas in which visual responsivity decreases during running)!
Quote Tweet
My paper with @jpillowtime analyzing the Allen brain observatory Ca2+ imaging dataset is finally out! rdcu.be/cKaBM
#tweetprint to follow
Show this thread
1
9
77
just to illustrate some example true and false transients:
2
Show this thread