How do RNNs encode and maintain high dimensional visual information to perform memory-dependent tasks? .Checkout the new paper led by PhD student @xiaoxuan_lei (NeurIPS2024) with @taku_ito1.

RT @fenildoshi009: 🧵 What if two images have the same local parts but represent different global shapes purely through part arrangement? Hu….

Come up north!.

🚨New paper 🚨.World models bake knowledge into weights, making them slow to adapt when environments change. We built one that infers from episodic memories instead. Check it out if you are interested in world modeling, episodic memory, spatial reasoning, or all of the above.

RT @dyamins: Come to our CCN workshop! Blending iophysical constraints and neural networks. Topographical ANNs + and much more. @meen….

Come by @great_caster poster if you are at #ICLR2025 and interested in topographical neural networks! .When📅: Friday, April 25 from 10:00 AM to 12:30 PM Where: Hall 3 + Hall 2B #60.

At #COSYNE2025 ? Come by @MotaharehPR 's poster at 1pm [2-089] if you like to know about how cool her NN model of visual search is! Trained de novo on visual search, its fast, efficient, mimics human behaviour, and has a beautiful neural geometry.

RT @pcastr: Looking to hire a student researcher to work on cool project for 6 months in DeepMind Montreal. Reqs: .- Full-time masters/Ph….

RT @roydanroy: Very troubling. I believe the U.S. has just ruled itself out as a reasonable destination for scientific conferences. @NeurI….

Check out our latest work (and the tweeprint), Credit-based SOMs for training topographical deep nets! Very proud of this work led by the first undergraduate in my lab! Great job @great_caster.

If you’re at NeurIPS, Check out @xiaoxuan_lei poster on representational geometry in RNN models of working memory using natural sensory inputs!.

RT @ShahabBakht: This is a great opportunity to secure funding for postdoc. My lab has a few research directions that align with the IVAD….

RT @jpineau1: Interesting analysis on the impact of open LLMs.

4/ simulating neural activity with linear mapping methods are important tools for both our scientific understanding and pushing towards potential applications like neural control.

3/ linear regression is not the only way for finding linear maps. Constrained domain specific approaches like those from @bethgelab and @AToliasLab @sinzlab generalize much better and should be more widely used/studied.

2/ but often the observed overfitting is due to lack of sample diversity and larger datasets like @KendrickKay NSD and continuously growing BrainScore are important steps towards more reliable ranking for models with these methods.

1/ I agree that linear regression is not a silver bullet. We already know that simple regression-based maps don’t generalize well to OOD.

I have been following the discussions around new work from @RylanSchaeffer et al. in the past few days including very thoughtful responses by @aran_nayebi @AndrewLampinen that pointed out several misconceptions. A few points worth adding:.

RT @Mila_Quebec: Mila's annual supervision request process opens on October 15 to receive MSc and PhD applications for Fall 2025 admission!….

RT @ShahabBakht: I’ll be recruiting 1-2 graduate students for Fall 2025 to work on visual learning generalization in humans and artificial….