4 ms·
If anyone's interested in some technical details: - The algorithm is kind of like collaborative filtering, but it filters over previous listening sessions of t
by jacobobryant 7y ago
If anyone's interested in some technical details:
- The algorithm is kind of like collaborative filtering, but it filters over previous listening sessions of the same user rather than different users.
- Backend is written with Clojure on Datomic cloud ions. Having Datalog queries for one thing was a huge help. Not sure if the performance characteristics of Datomic will make it work for this use case in the long term, but I'm planning to stick with it until I have to move.
Sorry the installation procedure is a little janky. I was going to polish it but decided it would be a better use of time to move immediately into developing mobile apps instead since those will reach a wider audience. Let me know if you run into any install issues though.
I've been using this for ~1.5 weeks now and it's been really nice. I hope someone else finds it useful.
- dustingetz 7y agoVery cool Jacob!
- jacobobryant 7y agoThanks Dustin :) I was interested to see this other post show up on HN yesterday too: https://news.ycombinator.com/item?id=20574182#20583543 https://news.ycombinator.com/item?id=20574182#20583543 At this early stage, I haven't started using any matrix factorization techniques since scaling isn't a huge issue when you only have two active users, but it'll be very relevant later on.
- thom 7y agoWould be fascinated to see a writeup on Dataomic being used in anger.
- jacobobryant 7y agoI have a writeup for a previous app I made (a budgeting app) here: https://jacobobryant.com/post/2019/ion/ https://jacobobryant.com/post/2019/ion/ What things would you be interested in reading about? I'll likely have enough material for another writeup soon.