After 1.5 years of hard work, I am thrilled to share with you Φ-SO - a Physical Symbolic Optimization package that uses deep reinforcement learning to discover physical laws from data . Here is Φ-SO discovering the analytical expression of a damped harmonic oscillator👇 [1/6]

Typst has launched its beta! Register now on typst.app!

And now translated in English : The Hartree-Fock method illustrated on 1s STO-3G orbitals with a restrained formalism. Fully coded in python, with all the equations used in Szabo given #compchem github.com/MartinVerot/Ha…

Three years ago, I started working on an easy-to-use tool for interpretable machine learning in science. I wanted it to do for symbolic regression what Theano did for deep learning. Today, I am beyond excited to share with you the paper describing it! arxiv.org/abs/2305.01582 1.

On Thursday I'll be giving a remote talk at the Rust Edinburgh meetup 🎙️ I'll be covering Rustdoc's JSON format, a machine-parsable format to analyze the public API of Rust crates. Since it's a remote talk, everything will be streamed—mark the date and time!

At Mainmatter we believe that Rust has incredible potential in the backend, but it still takes courage to make the jump. Starting today, I'm opening up my calendar: if you're considering Rust but you are not quite sure it's a good fit, come and have a chat!

Just published our latest work on pre-Born-Oppenheimer wave functions! Actually, in a pure state, all molecules are spherically symmetric! We showed how to recover the spherical symmetry using projection operators. pubs.acs.org/doi/10.1021/ac…

Curious about 2RDM methods? Want to write your own variational 2RDM code? Check out this tutorial review: arxiv.org/abs/2310.10746…

Very proud of our Neil Judge 🏴󠁧󠁢󠁳󠁣󠁴󠁿🇨🇭 for a stellar performance on the defence of his #PhD DCBPunibern The #scottish 🏴󠁧󠁢󠁳󠁣󠁴󠁿 party afterwards will go down in the history books! Thanks so much AlbrechtResearch and Mike Hill Chemistry@Bath Uni for taking part on the examination

Catching on ThePrimeagen's stream from yesterday on Mojo vs Rust: he's right. If Mojo delivers, we'll never see Rust in "userspace" for people working on AI. The value proposition is just *too good* if you're an ML engineer or Data scientist who knows Python well.

Diffusion models can be reformulated as equilibrium thermodynamical systems.

MLPs are so foundational, but are there alternatives? MLPs place activation functions on neurons, but can we instead place (learnable) activation functions on weights? Yes, we KAN! We propose Kolmogorov-Arnold Networks (KAN), which are more accurate and interpretable than MLPs.🧵

Wow this blows me away you can get roughly correct structure of pure electrolyte solutions out of this.

i wouldn't say the productivity boost is a raw one. its not because i am literally operating faster because of the time savings of keeping your hand on the keyboard and the efficient key binds. but instead it feels a lot better to code for long hours

Our entire product is powered by these 30 lines of boring Rails code

This book is a masterpiece! What I really like about it is that it teaches you the concepts and ideas, shows you some pseudocode, and leaves the implementation to you in whatever language you prefer! nostarch.com/writing-c-comp…

Oldies but goldies: James Sethian, A Fast Marching Level Set Method for Monotonically Advancing Fronts, 1996. Proposes an n*log(n) algorithm to compute the geodesic distance discretized on a grid by modifying Dijkstra's algorithm. pnas.org/content/pnas/9…

"Tensors from Scratch" - I am publishing the first part of the blog series that will help a beginner understand and implement tensors in C from scratch, starting all the way from defining tensors and shapes!

A 150 page review paper on the applications of machine learning in finance. #machinelearning #finance

Dreams are synthetic data. But why use data so unconnected from the real world? My answer is that it probably helps with overfitting to our daily lives.