deppep
Joined 2 years ago
Karma
23
Submissions
26
Member since
Feb 15
Recent submissions
Comment · 10 days ago
once again, back to firefox
Comment · 2 months ago
UMAP plot for a word2vec recommender over ~17M between tracks, albums, artists and labels. Try moving around.
Comment · 3 months ago
missing a lot of amazing federician castles in Apulia, Italy
Comment · 5 months ago
i don’t really see it this way. the value of a token in Python is much higher than it is in lower-level language
Comment · 5 months ago
Thank you for the head-up!
Comment · 5 months ago
thank you! :)
Comment · 5 months ago
i also made something like this. it cover 17M entities across tracks albums artists and labels. posted on show hn a few times but it went unnoticed (hate u (joking))
https://toposonico.com/#lon=14.4313&lat=-1.0200&z=9.10&entit...
Comment · 5 months ago
Hey. Needed this for a while, maybe some other hn people into astrophysics will find it useful too. It's a light package for parsing NASA GCN notices into Python data struct. Parsed notices are flat, fully typed and named so it comes easy to work with them with the autocompletion and static analyzer from your favourite IDE.
That's it, not the funniest project for sure but hope that's useful to someone.
repo: https://github.com/peppedilillo/gcn-notice-parser docs: https://peppedilillo.github.io/gcn-notice-parser/
Comment · 5 months ago
Hello! I find these fascinating too. The 128 dimensions I mentioned are learned by the NN, they are latent space and their actual meaning is ultimately unknown. The number itself is somewhat arbitrary as it was selected optimizing memory usage vs data coverage so that the track embeddings could sit on an A100.
Comment · 8 months ago
lovely. how was the visualization made?