Jeffrey C. Erlich @jerlich@qoto.org
@eikofried A step in the right direction, but not far enough. I tweet shamed the Guardian writer who wrote the piece. I guess they got quite a bit of criticism!
Today!!! Go visit 567.08 / UU14 - The rat frontal cortex encodes a value map in support of economic decisions under risk!
https://www.abstractsonline.com/pp8/#!/10619/presentation/69748
related preprint: https://www.biorxiv.org/content/10.1101/2021.11.19.469107v2
Jeffrey C. Erlich
boosted
🚨 BIG DATA RELEASE 🚨 We are beyond excited to announce the release of our Brain Wide Map of neural activity during decision making! It consists of 547 Neuropixel recordings of 32784 neurons across 194 regions of the mouse brain 🐭🧠
All these recordings were performed in a distributed fashion in 12 different labs, spanning Europe and the US 🌎 Rigorous standardization of methods and materials allowed us to pool the data from these labs together into a single gigantic dataset 🐙
Mice are performing our standardized perceptual decision-making task in which they have to position a stimulus in the center of a screen to receive reward. The dataset contains the stimuli and decisions, but also videos from three angles and DeepLabCut pose information. We're even releasing all the raw ephys data!
We know, it's a lot. At your own pace you can read all the details about the experimental setup, the task, processing of the data, and much more in the technical paper which accompanies this data release: https://figshare.com/articles/preprint/Data_release_-_Brainwide_map_-_Q4_2022/21400815
To explore the data at your leisure, visit our visualization website where you can scroll through different recording sessions, look at neural activity during example trials, and see trial-based activity of single neurons: https://viz.internationalbrainlab.org
Do you have itchy fingers to run your models on this humongous dataset? We totally get it! Here you can find how to download the data using our API so you can fire up those computing clusters: https://int-brain-lab.github.io/iblenv/notebooks_external/data_release_brainwidemap.html
This was a collective effort of our stellar team, who all put in so much work to make this monumental achievement possible. Our collaboration consists of 22 PIs, 37 researchers, and 11 staff members who all worked tirelessly to bring these data to you, the community 👏🍾
Jeffrey C. Erlich
boosted
The lab has two posters at SFN this year!
567.08. The rat frontal cortex encodes a value map in support of economic decisions under risk https://www.abstractsonline.com/pp8/#!/10619/presentation/69748
217.11. Distinct roles for rat premotor cortex in World-Centered versus Self-Centered Planning
https://www.abstractsonline.com/pp8/#!/10619/presentation/83705
Jeffrey C. Erlich
boosted
Question: how much flexibility do instances have to add/play with features (e.g. the mechanics of threading, something like QT, etc)? Is it worth thinking about the features that facilitate scientific discussion, the needs of our community, and how they might be implemented?
@barefootstache I think industry sponsored postdocs is definitely one way mechanism that might be sustainable.
The number of grad students who want to stay in academia is dropping because dramatically better salaries and work/life balance compared in industry
Although this trend makes my life harder (since it is very hard to recruit good postdocs), i see it as generally positive. We need to be thinking about new sustainable models for conducting research in academia - and this "tipping point" will force us to grapple with this.
Jeffrey C. Erlich
boosted
We have sooooo many tutorials :-). If you don't mind sitting down for a longer read, this is a community-written guide that goes into a lot of detail:
https://github.com/joyeusenoelle/GuideToMastodon#an-increasingly-less-brief-guide-to-mastodon
Jeffrey C. Erlich
boosted
🇳🇱
I made my qoto account months ago, but haven’t really spent time in the fediverse until this week. Good to be considerate to those who have built a community and culture here for years.
Jeffrey C. Erlich
boosted
My first post on Mastodon: please take part in The Perception Census - help us advance research into perceptual diversity, and learn about your own powers of perception too: https://perceptioncensus.dreamachine.world/
Jeffrey C. Erlich
boosted
An #introduction to eLife's new Mastodon page!
We're an #OpenAccess not-for-profit journal that publishes and reviews #research in the life and biomedical sciences.
We want to improve the way research is practised and shared in part by working with early-career researchers #ECR and supporting #OpenSource technology.
We also just announced our new publishing model that we hope will tackle an overreliance on journal titles and publishing decisions as quality measures for science and scientists.
@eikofried Great critique. Seems the review process failed here. If I was reviewing that paper I would not have allowed them to keep the title. And shame on the Guardian for boosting such a weak finding.
Jeffrey C. Erlich
boosted
In case someone didn't know, two books I've co-authored are freely available online for non-commercial use:
#Bayesian Data Analysis, 3rd ed (aka BDA3) at https://stat.columbia.edu/~gelman/book/ and lectures plus #rstats, #Python and #Stan code at https://avehtari.github.io/BDA_course_Aalto/
#Regression and Other Stories at https://avehtari.github.io/ROS-Examples/ including #rstats and #Stan code
The web sites also have links to the publishers' web stores if you prefer hard copies of these
The [NYTimes published an article on which steel plants make the most pollution](https://www.nytimes.com/2022/11/09/climate/climate-change-emissions-satellites.html) and it looks like China is by far the worst. But this map doesn't tell the whole story. Who is buying and using that steel from China? If a substantial portion of that steel is sold to the west then that pollution should really be attributed to the client countries, not the manufacturing countries (or at least show both maps).
Jeffrey C. Erlich
boosted
Registration to “What is memory?” Brain Mind Institute Symposium is now open! Places are limited and it’s first come first serve so go for it, now, & please boost this news! Looking forward to seeing you all there!
We read this very cool paper from Ayelet Sarel, Shaked Palgi, Dan Blum, Johnatan Aljadeff, Liora Las & Nachum Ulanovsky in our lab meeting today.
https://www.nature.com/articles/s41586-022-05112-2 (open access!)
They recorded from two bats flying back and forth in a 135m tunnel. They wirelessly recorded from the hippocampus of one of the bats and recorded the position and ultrasonic vocalizations of the bats. As the bats approach each other, they dramatically increase their vocalizations (presumably to avoid collision).
Hippocampal place cells are modulated by the relative position of the animals in complex ways, and this modulation is very rapid (turns on and off within a few seconds) as the bats approach each other.
I think the paper is quite a beautiful demonstration of how we can use ethological phenomena (like the sudden appearance of a conspecific) to better understand neural dynamics.
That said, I have a substantial concern with the paper. Hopefully, this toot might find its way to the authors - encourage them to join mastodon and reply :)
The authors analyze the neural activity relative to the tunnel , which they call "position" and with respect to the other bat which they call "interbat distance". However, the measure that they call "interbat distance" is not distance. Euclidean distance along a line is defined as +√((x₂-x₁)²). **It is always positive**. The authors _redefined_ distance to be the signed value x₂-x₁. This quantify is the position of x₂ relative to x₁, not the distance.
You might think that I'm being pedantic. Maybe the authors just thought it would be clearer to talk about "position" and "distance" instead of saying "tunnel position" and "position relative to other bat". However, they claim that "hippocampal neurons can rapidly switch their core computation to represent the relevant behavioural variables." If we change their wording to my wording, then what they have shown is that hippocamal neurons can rapidly switch the reference point that is being used to represent their current position, including using a reference point that is moving.
This is still a cool result. But is less novel. There is extensive work examining how hippocampus remaps or "reregisters" place cells when animals have to monitor multiple reference frames. A nice example is from [André Fenton's lab](https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1000403). The authors cite Fenton's work but say that Fenton "reported switching between two position maps, whereas here we found switching from position representation to distance-by-position representation." And that, i think really underscores how their redefinition of distance influenced the way they think about their results.
Be cautious in interpreting the results from fitting cognitive/computational models!
This is, in some ways, obviously true. But it is quite common in the field (and even in my own work) to "over-interpret" results from model fitting. I think the real long term solution to this is pre-registration, so that we are not using models to "discover" effects but to make specific predictions.
Cool new paper from Zador Lab:
