Be a content creator for the Neuromatch NeuroAI course! We're looking for people to write tutorials on transfer learning and RL in PyTorch over the next 3 weeks. If you want to help build this amazing course, DM me for details or fill out this app: airtable.com/app32npl2ZlbJv…

Cosyne Workshop Alert! On Tuesday March 5th, Chethan Pandarinath and I are proud to bring you:
Understanding Neural Computation using Task-trained and Data-trained Networks.
youtu.be/bJ0stLORdgQ

Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.

Today, I am very proud share what we have been working on for the last 14 months. ✨

Introducing Aya -- a new state-of-art for massively multilingual models. 🔥🎉

'Why the simplest explanation isn’t always the best' - commentary with Eva Dyer highlighting how dimensionality reduction does not usually give us what we want. pnas.org/doi/10.1073/pn… 'Major Achievement: Dino scatterplot in paper' unlocked.

How can we extract insights from behavior modulated by multiple complex factors? 🐭🪰🤖🏃

Check out our #NeurIPS2023 Spotlight Paper where we present a SSL method for learning multiscale representations of behavior! Machine Learning at Georgia Tech Google DeepMind

Link: multiscale-behavior.github.io 🧵

Neural spiking data and Transformers are a tricky match. Temporal segmentation and tokenization are the crux. Together with an all-star team, we figured out a scalable sol. The results are exciting: training and transferring on multi-sessions, multi-subjects neural decoding tasks

This is incredibly important. We will all be using such methods soon.

The key to building a universal encoder for spikes? Spike embeddings! Mehdi and co. build a single model that can train on any neural data that mapping spikes to a continuous space, similar to word embeddings. It works 🔥

Check out this new paper:

Led by Mehdi Azabou and Eva Dyer, we show that it is possible to get SOTA brain decoding with transfer across individuals and tasks!

The key is a clever way to tokenize spiking data for transformers.

#brain #neurotech #NeurIPS2023

Super excited to share new work to appear at #NeurIPS2023 led by the incredible Mehdi Azabou! In collaboration with an amazing team: Vinam Arora, 🤖, Ximeng Mao, Santosh Nachimuthu, Michael Mendelson, Blake Richards, Matt Perich, Guillaume Lajoie Patrick Mineault Machine Learning at Georgia Tech 🧠🤖

Is a universal brain decoder possible? Can we train a decoding system that easily transfers to new individuals/tasks?

Check out our #NeurIPS2023 paper where we show that it’s possible to transfer from a large pretrained model to achieve SOTA 🧠!

Link: poyo-brain.github.io 🧵

📢 Exciting News! Join us at the 𝐒𝐞𝐜𝐨𝐧𝐝 𝐖𝐨𝐫𝐤𝐬𝐡𝐨𝐩 𝐨𝐧 𝐒𝐲𝐦𝐦𝐞𝐭𝐫𝐲, 𝐈𝐧𝐯𝐚𝐫𝐢𝐚𝐧𝐜𝐞, 𝐚𝐧𝐝 𝐍𝐞𝐮𝐫𝐚𝐥 𝐑𝐞𝐩𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧𝐬, co-organized by Arianna Di Bernardo and myself, at #BernsteinConference 2023. 🎉

With a slight delay: 📸 in action at our Half-Hop poster #ICML2023 , featuring three of my awesome co-authors: Mehdi Azabou (lead author), Eva Dyer & Michal Valko ✨

(along with the watchful eyes of Michael Galkin 👀)

After intense days at #ICML2023 , time to slow down... literally! 🐌

Today, Mehdi Azabou presents Half-Hop; I'll also be there!

For each edge, w/ probability p, insert a node in-between, slowing message passing down in its tracks.🚦

10:30am--noon; Poster #407. Come say hi! 🦥

In a time where everything seems to be speeding up… Half-Hop improves learning by slowing down!

Learn how our simple yet powerful graph upsampling augmentation enhances learning in MPNNs! #ICML2023 🏝️

Paper: openreview.net/forum?id=lXczF…
Today 10:30am--noon; Poster #407

#ICML2023 's main program is coming down the home stretch. Here are two research options we think you might like for your Thursday agenda!

Check out the week's Georgia Tech highlights: sites.gatech.edu/icml-2023/high…

Morning #ICML2023 ! Check out some of our students presenting today. Poster Session highlights from Georgia Tech.

More at sites.gatech.edu/icml-2023/rese…

As ICML Conference kicks off its main program this week, read about a breakthrough in #ML for graphs from Eva Dyer's research group:

New Research from Georgia Tech and Google DeepMind Shows How to Slow Down Graph-Based Networks to Boost Their Performance

sites.gatech.edu/icml-2023/new-…

Machine learning experts at Georgia Tech share their latest breakthroughs next week at ICML Conference. Meet some of the people shaping the #AI field.

Learn about their work and what #AI challenges and opportunities they see on the horizon.
sites.gatech.edu/icml-2023/