In case someone is looking for historical weather data for ML training and prediction, I created an open-source weather API which continuously archives weather data.
Using past and forecast data from multiple numerical weather models can be combined using ML to achieve better forecast skill than any individual model. Because each model is physically bound, the resulting ML model should be stable.
Judging by the URL, this book was used for CMU's 15-151 / 21-128, which is a first-semester course for CS and math undergrads. Nowadays, the course uses [0].
I think this is a really good primer for electronic music production if you're going to start from absolutely zero.
If you specifically want to program beats, I recommend this quirky book "Pocket Operations": https://shittyrecording.studio. It's basically guitar tabs for drum machines. Pick out some styles as a foundation and then build on top of it. Think of it as boilerplate code.
Being a software dev by day, and a former musician in high school, the current world of digital music production tools is as incredible as it is overwhelming. It's good to have something that orients your practice and experimentation.
There was an amusing study where they tried to measure how phonemic various language orthographies are by training an ANN model on some data to convert back and forth between spelling and pronunciation, and then checking the error rate on more inputs: https://arxiv.org/abs/1912.13321. Here's the table from that paper; numbers are percentages of accurate guesses.
It's easy to see here that English is truly dismal, so being better than it is not a great achievement. Spanish is okay but far from the best. Esperanto (eo) tops the chart for obvious reasons, but note how Serbo-Croatian (sh) is an example of truly outstanding phonemic orthography for an already existing language. Turkish (tr) and Finnish (fi) are also impressive.
I self-host literally everything (email, calendar/contacts, VOIP, XMPP, you name it) from by basement with used 1U servers from eBay and a cable internet connection.
It was probably more hassle than most people would want to bother with to get it set up. But, with everything up and running, there's very little maintenance. I probably spend a few hours a month tinkering still, just because I enjoy it.
I use a stack of Proxmox VMs, FreeIPA for authn/authz, and Rocky Linux for all servers and workstations. My phone runs GrapheneOS with a Wireguard VPN back to the house. I don't expose anything to the public internet unless absolutely necessary.
I recently anonymized and Ansibilized my entire setup so that others might get some use out of it:
I'm in the deep learning music scene, which is due for its stable diffusion moment in the next year or two. The (primarily) timbre transfer system called RAVE is where I'm starting, and my contribution is to optimize the system to improve training time.
Didn't expect to see sewing patterns on HN, but not complaining, this looks really cool!
While on the topic, my girlfriend recently showed me this site: https://freesewing.org . It's fully open source and all the patterns are actually parametric, meaning you can customize the patterns to exact body measurements. Their custom JS framework for designing patters as well as the rest of the platform is MIT licensed [0] and the patterns are all Creative Commons.
It's a really good pixel art tool, but I think their licensing strategy is particularly interesting to the hn audience. The tool is open source, but they don't distribute binaries for free. So if you just want to use it, you can buy a copy like normal software, but if you've got the technical ability to do so, you could just build a binary yourself from source. Always struck me as such a good compromise.
I have a gallery of split keyboards, including this one. It's useful if you need an overview of what's available.
I have an ErgoDash with a Dvorak layout and a 3D printed adjustable tenting stand. The Djinn is fairly similar, though the additions are features I don't care for (encoders, screens, LEDs).
I've bought most of the parts to make a Lagrange, but still need to order the PCBs. I expect this to be an improvement on the ErgoDash, since I can include the missing keys (F1-F12 etc).
Just buying a Kinesis Advantage2 would probably have been a whole lot less effort.
My dream drawing app would let you draw a stroke multiple times and then pick out your best attempt. No need to undo all the bad ones to try again. It separates the generation and the judgment.
> growing crystals was of paramount importance for certain structure determinations using X-ray crystallography (to answer "did I really prepare what I think I prepared?")
Ahh, X-ray crystallography, a.k.a. the field of modern (bio)chemistry that most closely resembles medieval alchemy.
The excellent 2009 documentary "Naturally Obsessed" [0] is about how difficult, time-consuming, and frankly irrational it can be to successfully crystallize certain proteins and protein complexes. The film is about how some PhD students spend the entire five-ish years of their degree programs trying (and mostly failing) to find the secret recipes (protein purity/additive ingredients/temperature/humidity/phase of the moon) that will allow their proteins to crystallize, and thus granting them data with which to populate their dissertations.
Why irrational? In one case, the secret ingredient is pickle juice.
Using past and forecast data from multiple numerical weather models can be combined using ML to achieve better forecast skill than any individual model. Because each model is physically bound, the resulting ML model should be stable.
See: https://open-meteo.com