nivter

The U.S. government has said that all federal agencies will use quantum-resistant cryptography by 2035

I wonder what the current pace and effort of adopting this new cybersecurity paradigm is, and how familiar (AI) agents are with it. And I would assume most countries actively working on it must have amassed encrypted data, not just China and Russia.

Side note: I found the term quantum-resistant instead of quantum-proof a bit amusing. Also it would be better if the article included some references.

Some use cases and examples are provided here: https://www.tinyvolt.com/blog/bringing-autodiff-to-geometry

Not a fan of subscription hell myself, I plan to use one-time payment for all my products. Implementation wise one time payment is much simpler than setting up smaller payments that cap at a fixed amount.

Automatic differentiation. For any DAG with a scalar output, it allows calculating its partial derivative wrt the input parameters.

Youtube: https://www.youtube.com/@MITHANLab Course: https://hanlab.mit.edu/courses/2024-fall-65940

This is far from what I expected. There is not much related to quantization, pruning, common architectures, precision or benchmarking. For those interested in this topic, I would recommend content from MIT HAN Lab.

I believe that abstraction is recursive in nature which creates multiple layers of abstract ideas leading to new areas or insights. For instance our understanding of continuity and limit led to calculus, which when tied to the (abstract) idea of linearity led to the idea of linear operator which explains various phenomena in the real world surprisingly well.

If it's about teaching and not about efficiency, why not just use plain Python? One could argue it is actually better since students don't have to worry about typing and syntax, and it allows a gentler introduction to commonly used tools like jax and numpy while getting comfortable with the language.

A journey from rotation matrices to Lie algebra

Thanks, this is a very helpful feedback.

This was a joyful epiphany for me when I encountered continuous linear systems for the first time. Another corollary of this is when the kernel is δ(x-y), the resulting integral has the value f(x). I like to see it as a continuous version of expressing a vector as a sum of its components.

Yup, I consider this interpretation as a matrix being a function that takes in objects like a line, a circle or a convex shape and spits out objects like some other line, an ellipse or another convex shape. It is a level of abstraction where you no longer care _how_ matrix multiplication works - you mostly care about what a matrix does to geometric objects. I covered this aspect not in the above article but in a separate one: https://www.linearalgebraforprogrammers.com/la/3_mat_vec_mul

Originally I had planned to add a line or two explaining how a matrix-matrix product is just a _list_ of matrix-vector products but then dropped the idea to keep it solely focused on the interpretations. I will probably add it to make it all about matrix-matrix products.

Excited to share something I have been working for over a year in my spare time - an e-book of visual essays titled "Linear Algebra for Programmers" - https://www.linearalgebraforprogrammers.com/ (not optimized for mobile devices yet)

It starts with taking a weighted sum of numbers (and then vectors/arrays). Everything else builds up by just observing and interacting with weighted sums of arrays.

Hope you enjoy it.