My notes and rants.
People have been playing around with the latest open-weight video model MiniMax H3 (Hailuo 3) since its release. ComfyUI shipped day-0 support in this blog article , and that's when everybody started trying to run the model on their own computers—including me. However, for a MacBook user, the PyTorch support for Apple's MPS backend is not really that great. The MiniMax H3 workflow happens to step…
I just had my 24th birthday yesterday. Since it has been a very bumpy ride this past year, I am very willing to self-reflect and give an honest review of lessons learned from my past year and what I should consider improving. My 23rd year in life happens to sit in the context of a very bizarre historical window. The current US president is respectfully a warmonger and insider (pronounced “public”)…
Structure prediction asks what shape is compatible with a sequence. Protein design often asks the reverse question: which sequences are compatible with a desired shape? This reverse problem is called inverse folding . It sounds easier because the backbone is already supplied, but the answer is not unique. Many sequences can adopt similar folds, while small changes at a tightly packed core or…
LaTeX is powerful, but writing it becomes tiring very quickly once a formula gets complicated. Most of the time, we end up spending more time jumping between braces than thinking about the formula itself. For years, I dealt with this by moving my writing between several applications. Each one solved part of the problem, but none solved the whole thing. The LaTeX Suite desktop app came from wanting…
Calling proteins a language is useful only if the analogy is handled carefully. Amino acids are not words with dictionary meanings, and evolution does not compose sentences with an intention. A protein sequence is a physical polymer that must be synthesized, folded, and selected in an organism. Yet sequences do have context-dependent regularities. Some residue combinations are common, some are…
By the late 2010s, protein structure prediction was a chain of strong but distinct procedures. A sequence search found relatives. An alignment summarized their variation. A neural network predicted contacts, distances, and orientations. An optimizer then searched for coordinates that satisfied those predictions. Each stage had been improved, but information moved mainly in one direction. The…
Imagine comparing the same sentence as it has been copied and edited in many communities for millions of years. Some words change freely. Others remain nearly fixed because changing them destroys the meaning. A third group changes in pairs: when one word is replaced, another tends to change as well because the two must remain compatible. Protein families leave a record with a related structure.…
Recently I’ve been working on predictive and generative models of proteins. Quite a shift in expertise, if you ask me! Therefore, I may as well write some literature review posts about proteins here. (Not that I have no more SWE knowledge to share, and not that AI has fully replaced us for non-cutting-edge material to write about. Well... 🙁) P.S. I’m sorry for not updating anything since March;…
This post is a rant. A thought-provoking rant, I hope so. No, I don’t mean Luddism in a pejorative sense. I am referring to the Luddite movement in England, which took place in the early 19th century. And no, I don’t plan to reiterate the well-worn debate about how AI replacing humans is justified / unjustified. The world is not black-and-white, nor is it something we can resolve through political…
The following exercises are from Chapter 2 of Quantum Computation and Quantum Information by Isaac Chuang & Michael Nielsen. Cauchy–Schwarz inequality: | ⟨ v ∣ w ⟩ | 2 ≤ ⟨ v ∣ v ⟩ ⟨ w ∣ w ⟩ for any two vectors | v ⟩ , | w ⟩ . Proof 1 : Construct | z ⟩ = | v ⟩ − λ | w ⟩ , where λ is any scalar in C . Then 0 ≤ ⟨ z ∣ z ⟩ = ⟨ v ∣ v ⟩ − λ * ⟨ w ∣ v ⟩ − λ ⟨ v ∣ w ⟩ + | λ | 2 ⟨ w ∣ w ⟩ . Let λ = ⟨ w ∣ v…