mayukhdeb

7 ms·

1.
by mayukhdeb 2y ago
Thanks for the kind words! Happy to know that there are people out there who find this stuff just as interesting as I do.
2.
by mayukhdeb 2y ago
In this paper, we don't zero out the weights. We remove them.
3.
by mayukhdeb 2y ago
> Is it that the structure clustered the neurons in such a way that they didn't need to be weighted Yep. Because of the structure, we did not have to compute the output of each weight column and simply copied the outputs of nearby w
4.
by mayukhdeb 2y ago
Thank you for sharing this! We'll read through this and update the camera-ready version accordingly for ICLR 2025.
5.
by mayukhdeb 2y ago
Thank you for your kind words!
6.
by mayukhdeb 2y ago
Thank you for your kind words! Indeed. The problem with most AI research today is they simply do trial and error with large amounts of compute. No room for taking inspiration from nature, which is requires more thought and less FLOPS.
7.
by mayukhdeb 2y ago
Thanks for clarifying your reason for renaming the title. The explanation for the original title is this plot from our publication in ICLR 2025: https://toponets.github.io/webpage_assets/FigureEfficiencyNa... You can f
8.
by mayukhdeb 2y ago
If by popular fantasy you mean replicating the functional profiles of the visual and language cortex of the brain, then yes. These ideas in neuroscience are popular, but not fantasy. I encourage you to read up on functional organization in
9.
by mayukhdeb 2y ago
> the features tend to have greater semantical overlap? This is true. The features closer together now have much stronger semantic overlap. You can watch how the weights self-organize in a GPT here: https://toponets.github.io
10.
by mayukhdeb 2y ago
It is indeed brain-like in a functional way. Topographic structure is what enables the brain to have low dimensionality and metabolic efficiency. We find that inducing such structure in neural nets made them have significantly lower dimensi
11.
by mayukhdeb 2y ago
We localized "toxic" neurons by contrasting the activations of each neuron for toxic v/s normal texts. It's a method inspired by old-school neuroscience.
12.
by mayukhdeb 2y ago
Yep. That is exactly the idea here. Our compression method is super duper naive. We literally keep every n-th weight column and discard the rest. Turns out that even after getting rid of 80% of the weight columns in this way, we were able t
13.
by mayukhdeb 2y ago
Indeed. What's cool is that we were able to localize literal "regions" in the GPTs which encoded toxic concepts related to racism, politics, etc. A similar video can be found here: https://toponets.github.io More
14.
by mayukhdeb 2y ago
The motivation was to induce structure in the weights of neural nets and see if the functional organization that emerges aligns with that of the brain or not. Turns out, it does -- both for vision and language. The gains in parameter effici
15.
by mayukhdeb 2y ago
> The only potential benefit Other benefits: 1. Significantly lower dimensionality of internal representations 2. More interpretable (see: https://toponets.github.io ) > 7B model down to 6B We remove ~80% of the parameters
16.
by mayukhdeb 2y ago
Our goal was never to optimize for performance. There's a long standing hypothesis that topographic structure in the human brain leads to metabolic efficiency. Thanks to topography in ANNs, we were able to test out this hypothesis in a
17.
TopoNets: High performing vision and language models with brain-like topography(arxiv.org)
225 points by mayukhdeb 2y ago|68 comments