Like many friends and fellow scientists, I am leaving this place and will move over to @bluesky. I realized around 2/3 of the people I’m following are already there. If you’re interested in our work, follow me and find me there at @aecker.bsky.social — bsky.app/profile/aecker.bsky…
A Perspective from the Ecker lab discusses the progress and challenges of using computer vision approaches for behavior studies of primates in natural environments.
nature.com/articles/s41592-0…
After 7 years, thrilled to finally share our #MICrONS functional connectomics results!
We recorded activity from ~75K neurons in visual cortex in a single mouse, then mapped its wiring using electron microscopy. To systematically characterize neuron function, we built the first foundation model of the mouse visual cortex—trained via deep learning on data pooled from multiple mice and visual cortical areas.
Our foundation model generalized to new neurons, animals, and even unseen stimulus domains. It also accurately predicted entirely new modalities, such as anatomically defined cell types. Importantly, this robust generalization enabled us to create accurate functional digital twins of individual mouse brains.
Using the digital twin of the MICrONS mouse—where we knew the exact neuronal wiring—we discovered that neurons don’t connect randomly, even when anatomically positioned to do so. Instead, given multiple potential partners (axons near dendrites), neurons preferentially choose partners with similar feature selectivity (“what”) rather than receptive field overlap (“where”).
Foundation models offer a powerful approach to systematically decode the neural code of intelligence.
Huge thanks to @IARPAnews for funding this groundbreaking effort through the @BRAINinitiative, and to our amazing team at @Stanford@StanfordMed@bcmhouston, @Allen, @Princeton, @uniGoettingen and others!
#Neuroscience#MICrONS#NeuroAI#Connectomics#FoundationModels#AInature.com/immersive/d42859-…
Poster – Fri 13 Dec 11 a.m.
Retrospective for the Dynamic Sensorium Competition for predicting large-scale mouse primary visual cortex activity from videos
neurips.cc/virtual/2024/post…
Also check out some our posters by the amazing @pollytur1 and many others from our collaborators @sinzlab and @AToliasLab:
Spotlight Poster – Wed 11 Dec 4:30 p.m. Reproducibility of predictive networks for mouse visual cortex
neurips.cc/virtual/2024/post…
Just arrived in Vancouver for #NeurIPS2024. Let's meet up if you're there. If you're looking for postdoc opportunities in #NeuroAI, touch base! We're always looking for talented postdocs.
🔍Why TreeLearn is important:
• Supports precision forestry for forest management & climate research🌍.
• Highlights an interesting CV problem 👁: Trees are complex, posing a challenge to current 3D instance segmentation methods that are usually evaluated on simpler scenes.
🛠️We release the TreeLearn code, model weights, and a benchmark dataset with evaluation code.
📊The benchmark dataset enables training and systematic comparison of existing tree segmentation algorithms.
📂Code: github.com/ecker-lab/TreeLea…
Forestry and computer vision researchers, meet TreeLearn🌲, a deep learning method for segmenting individual trees from forest point clouds.
⚙️It projects points toward tree bases & groups them via density-based clustering.
📂Paper: doi.org/10.1016/j.ecoinf.202…
I love it. Thanks eLife for taking a stance, continuing to move forward towards proper academic publishing and not getting distracted by antiquated business models that should have no place in science anyway. Way to go!
Following the news that eLife will not receive an Impact Factor in 2025, we’ve shared an update on how our model is doing since we were first placed “on hold” by Web of Science, and what we’re up to now. Find out more.
elifesciences.org/inside-eli…
On my way to Frankfurt for the #BernsteinConference. Let's meet up if you're there. If you're looking for postdoc opportunities in #NeuroAI, touch base. We're hiring!
To ask how similar the brain is to a neural network we need a similarity metric. In a new paper I asked how much the metric matters to downstream conclusions, and, upshot, it matters a great deal. biorxiv.org/content/10.1101/… (1/7)
We develop large-scale predictive models of the visual system in mice and monkeys. We use these models to understand the functional organization of the visual cortex and how function relates to other modalities such as morphology and connectivity.
We're hiring postdocs and PhD students in ML for neuroscience. Join our lab in Göttingen, Germany, and be part of an exciting international network!
eckerlab.org/applications/