Definition: E is evidence of A if P(A|E)>P(A) Proposition: absence of evidence is evidence of absence Proof: assume E is evidence of A. Then P(A)•(P(E)+P(-E))=P(A)=P(A|E)•P(E)+P(A|-E)•P(-E) Since P(A|E)>P(A), it follows that P(A|-E)<P(A) to compensate Hence P(-A|-E)>P(-A)

The paradox of poetry and information theory

You’re probably familiar with Fourier transforms. But, did you know that it can be viewed as a 90° rotation in the time-frequency domain and we can rotate by any angle via fractional Fourier transform. Explains why applying FT twice gives reflection (rotation by 180°).

Here it is! A huge milestone in neuroscience – the complete connectome of the fly brain from the FlyWire team (FlyWire). In this special issue, we have 9 research articles. I’ll try to summarize them in a couple threads starting here 🧵

neuroscience paper in 2045: "we recorded 5 million neurons in V1 with Neuropixels 9.0 probes while head-fixed mice watched Gabor flickers..."

This is a beautiful illustration of an apparent paradox in deep learning that ~the weights don't move~ I think we resolved this paradox in prior work, so I just want to share our perspective. And before you ask: yes, it's a question of norms 😅 (1/6) x.com/norabelrose/st…

Convergent transformations of visual representation in brains and models. arxiv.org/abs/2507.13941

Lots of discussion around JEPA and why latent space prediction works better than input space (e.g., LLMs) for certain modalities. But no one has formalized WHY. The answer lies in whether statistically dominant features are semantically meaningful. NeurIPS Conference spotlight 🧵👇

Here's an important question that we must contend with increasingly as AI-for-Mathematics pipelines become more commonplace: Why do we care about solving hard problems? Almost always, the answer is *not* because we particularly want the hard problem to be solved. (1/12)

They hard coded the variance - meaning the VAE encoder is only predicting the mean latent distribution, then they use a scaled identity covariance for the reparam trick - and bingo bango, no more instability and you can train everything e2e Salimans strikes again!

You: I wish I had a formula for this cursedly specific shape... Gielis Superformula: Yes!