←
John Beninger23 posts
John Beninger avatar

John Beninger

@jgbeninger

PhD candidate in computational neuroscience with @Neuronaud and @mossy_fibers, University of Ottawa. Mountain enthusiast.

◷ Joined July 2023▧ 6 Photos and videos
Richard Naud avatar
⟳ John Beninger retweeted
Richard Naud@NeuroNaud·28 Mar 2025

What are serotonin neurons trying to tell the brain? Our work on the question has been so much fun (@JCBeique , @JermiahYCohen, @grossmancooper ). Check out @efharkin_ 's exciting paper biorxiv.org/content/10.1101/… nature.com/articles/s41586-0…

Michael Lynn avatar
⟳ John Beninger retweeted
Michael Lynn@MichaelBLynn_·28 Jun 2024

Excited to share our updated preprint characterizing a novel, slow recurrent circuit motif in raphe! biorxiv.org/content/10.1101/…

Amir Zamir avatar
⟳ John Beninger retweeted
Amir Zamir@zamir_ar·17 Jun 2024

How far can a very simple eye go in solving vision tasks? Like a 1-pixel camera? Humans have one of the greatest eyes in nature, while many animals have significantly simpler eyes and visual systems yet show complex perceptual behavior. In an interesting project, we find that many computer vision tasks can be solved without a typical camera and with such simple 1-pixel sensors (photoreceptors). We also find that proper design (e.g., where to place the photoreceptors strategically) makes a big difference, so we developed a computational design method to find them. 🌐 visual-morphology.epfl.ch 👁️[Solving Vision Tasks with Simple Photoreceptors Instead of Cameras] 🧵1/n

0:39
OpenAI avatar
⟳ John Beninger retweeted
OpenAI@OpenAI·13 May 2024

Say hello to GPT-4o, our new flagship model which can reason across audio, vision, and text in real time: openai.com/index/hello-gpt-4… Text and image input rolling out today in API and ChatGPT with voice and video in the coming weeks.

1:22
Emerson F. Harkin avatar
⟳ John Beninger retweeted
Emerson F. Harkin@efharkin_·27 Feb 2024

Interested in serotonin and RL? Drop by poster 77 on Thursday night at #Cosyne2024 to hear about the connection between normative theory and biophysical implementation in the dorsal raphe. Come for the model, stay for benchmarks that blow CNN models of vision out of the water 💥

Post attachment
Emerson F. Harkin avatar
⟳ John Beninger retweeted
Emerson F. Harkin@efharkin_·24 Feb 2024

It's tough to adapt to a fast-changing world---unless you're a neuron. 📈 Build an intuition for current- and threshold-based spike frequency adaptation using the generalized integrate and fire neuron model of @neuronaud! 👇 efharkin.com/blog/2024-02-gi…

0:16
Zoza avatar
⟳ John Beninger retweeted
Zoza@ai_zoza·22 Feb 2024

We’re looking for design partners to help us shape the future of our product. If you’re potentially interested, please DM us or comment below!

Zoza avatar
⟳ John Beninger retweeted
Zoza@ai_zoza·22 Feb 2024

We’re super excited to announce the launch of Jamix (jamix.ai)! In addition, we’re proud to announce our $3M pre-seed round led by @AudaciousHQ with support from amazing angel investors! Read below for more details about our product.

John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

Are there precisely 5-6 subtypes of synaptic dynamics? What seems clear is that the dictionary of subtypes extends further than facilitation and depression. This added complexity hints at the possibility of multi-level (≥3) multiplexing or multiple high frequency coding schemes.

John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

This prediction only worked when classifying connections recorded in an extracellular calcium concentration of 1.3 mM. At 2 mM this prediction failed suggesting higher calcium may partly homogenize the synapses and obfuscate meaningful cell-identity specific dynamics. 9/10

Post attachment
John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

How important is this partial overlap of functional subtypes and transgenically defined identities? We performed supervised learning to predict transgenically defined identities using only the parameters of our model and found above chance prediction accuracy. 8/10

Post attachment
John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

We find that each of our functional subtypes includes multiple transgenically defined cell-types and each transgenically defined subtypes occupies multiple functional subtypes. Nonetheless transgenically defined cell types do preferentially inhabit particular functional subtypes.

Post attachment
John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

When theorizing about information processing in neural circuits we often assume that specific genetically defined cell types have specific STP properties and connectivity patterns. Do our functional subtypes have an underlying genetic basis? 6/10

John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

Clustering algorithms will always try their best to find something that looks like clustering so how can we know these clusters aren’t spurious? We tested our clustering pipeline on synthetic data corresponding to a continuum of STP properties and none explained our results: 5/10

Post attachment
John Beninger avatar
John Beninger@jgbeninger·19 Feb 2024

Clustering model parameters across hundreds of synapses, we found five functional subtypes of synaptic dynamics in rodents and six functional subtypes in humans which show qualitative overlap (though based on existing data we don’t expect a perfect overlap). 4/10

Post attachment