We had a fantastic lecture from Leo Kozachkov about how to connect #ML and the Brain. What a talk! Impressive results and unique insights. This will deffo fuel the next generation of #AI algorithms. Don't miss it:

youtu.be/vHGoBYjsG4c

Profound insights and super novel! Very…

We developed a biologically-plausible model of neuron-astrocyte networks, which in a certain limit describes a Dense Associative Memory with large storage capacity. This correspondence raises an intriguing possibility that memories might be stored, at least partially, in the…

I am a big fun of this work. The properties of energy landscapes in various models from the Dense Associative Memory family are still somewhat mysterious. This work is a great step towards their better characterization.

🚨Princeton ML Theory Summer School🚨

Aug 6 - 15, 2024

Speakers:

* F. Krzakala/L. Zdeborova Krzakala Florent Lenka Zdeborova
* S. Rakhlin
* B. Lourerio Bruno Loureiro
* M. Austern
* D. Krotov Dmitry Krotov
* J. Altschuler

Details: mlschool.princeton.edu

Sponsors:…

Associative Memory networks are mathematically different from the conventional feed-forward neural networks. It was fun to discuss the development of potential software tools and frameworks for this class of models.

Panel: Hopfield Networks meet Software Engineering at…

I started learning about astrocyte biology about 1.5 years ago. I find fascinating the idea that astrocytes can dynamically reconfigure synaptic weights of the neural network. It was fun to share what I have learned about these somewhat underappreciated cells with the NeuroAI…

I greatly enjoyed the talk by Carlo Lucibello on the storage capacity of Dense Associative Memories from the perspective of statistical physics. The connection to Derrida’s random energy model is very insightful!

Paper: arxiv.org/abs/2304.14964

I enjoyed being on the panel at #AMHN2023 , thanks for the great questions!

As promised here is a barebones HAMUX for those who want to get their hands dirty with deep Hopfield nets

1 file, <200 lines of (important) code, and a demo notebook

JAX ❤️

github.com/bhoov/barebone…

Gave a talk at the #AMHN2023 workshop at #NeurIPS2023 . It was a fantastic workshop on the rejuvenation of the Hopfield Network and its application in understanding computation in the brain and designing the next-generation of AI architectures.

“Let’s get rid of this Transformer nonsense” Sepp Hochreiter, NeurIPS 2023 Sepp Hochreiter

Spontaneous comments by Terry Sejnowski on the impact of Hopfield Networks on many subfields of AI.

“We, physicists think in terms of energy functions.”

The room is packed for John Hopfield ’s opening talk at the #AMHN2023 workshop at #NeurIPS2023 . His presentation covers the fascinating historical evolution of Associative Memory. Beginning with insights into the renowned NRP program at MIT, the origin of the term 'neuroscience”,…

Day 4 of Neurips: Today only water w/ Sepp Hochreiter Johannes Brandstetter Francesco Locatello Dmitry Krotov et al.

I am ✈️ to #NeurIPS2023 , and will be there until Saturday. Hit me up if you want to chat about science!

Here is a summary of our presentations:

- Long Sequence Hopfield Memory (Tuesday 10.45am): arxiv.org/abs/2306.04532. A network for storing temporal sequences with very…

Dense Associative Memory can store a large (possibly exponential) number of static high-dimensional patterns. What about sequences of patterns? For instance, when a songbird or a robot produces a sequence of actions (e.g., song vocalization) every state vector of the sequence…

I almost DDOS'ed myself by asking you for your NeurIPS papers! I found 197 NeuroAI and NeuroAI-adjacent papers at the main conference–brace yourselves! neuroai.science/p/neuroai-pape…