5 ms·
The parameters can be configured based upon system complexity etc. but there is always going to be some level of false positive. But that still may be okay if t
by bitsinthesky 4y ago
The parameters can be configured based upon system complexity etc. but there is always going to be some level of false positive. But that still may be okay if there are still those i x m x n lists posted around, one can quickly focus in on which lists are, let's say, 99.5% likely to be the relevant bus stop lists.
- elil17 4y agoSeems like the kind of thing that is great in theory but would end up with a high error rate in practice.
- pavel_lishin 4y agoIt also seems like it would actually be fairly difficult to explain to a majority of people, especially in a non-verbal way.
- michaelmior 4y ago> there is always going to be some level of false positive That's not necessarily true. Even if you're dealing with a large number of stops, routes generally aren't changing very frequently. This means you can afford to spend some time tweaking the output to get the false positive rate down to zero. This isn't strictly a Bloom filter anymore, but I think any level of false positive makes this a non-starter. I can't imagine any transit agency would knowingly to put up signs that will misdirect customers.
- infogulch 4y agoAh, a perfect bloom filter. https://www.researchgate.net/publication/260163952_Minimal_perfect_hashing_and_bloom_filters_made_practical https://www.researchgate.net/publication/260163952_Minimal_p... Low accuracy filters could still be useful if they are small enough, and paired with the raw table. Maybe you could get 93% true negative rate with just a 5x7 grid.
- michaelmior 4y agoMaybe, but I think educating transit riders on how to use them correctly would be a major challenge.