3 ms·
> We can do this once, and then binary search a safety interval around that position. Alas, this only gets rid of the fast jumps at the beginning of the binary
by leni536 4y ago
> We can do this once, and then binary search a safety interval around that position. Alas, this only gets rid of the fast jumps at the beginning of the binary search, and for some reason, it ends up being slightly slower than binary search alone.
I don't get why you can't do this at each step.
- genericlemon24 4y agoYou can, see the next paragraph, "also narrow down around the estimated position iteratively..." I gave up because I was too lazy to think how to do the "precompute the maximum +/- error bounds" jiggawatts mentions in https://news.ycombinator.com/item?id=34021826 https://news.ycombinator.com/item?id=34021826
- danielvf 4y agoBTW, loved the article, not sure how I missed that you tried this.