|
<< ctxmodel.net
I don't understand why people are so obsessed with using hashes, especially if that's the same people who care about speed so much. The only real hash advantage is simplicity - never the speed, even if its a small hash fitting into L1. So it'd be much better imho to forget about hashes and make some trees, though they're more complex to code ;) Well, entropy-based hashes are better than trees, but I doubt that anybody is already using that.
> too lazy to code tricky trees - actually i never thought > about using trees). One of my concepts doesn't work with > trees, a hash seeding function based on file type (allows > you to optimize hash/coding related stuff and to be able to > do specialized modelling). Actually I really like the straight approach recently. I mean, not designing the algorithms as incremental from the start. For example, if we have the data, its always possible to compute the necessary statistics at any place. Of course its slow, but allows to think about optimization after the simple and most general initial implementation. But this way its possible to choose data structures to make a known algorithm faster, and normally we're starting with data structures which don't allow much choice in algorithm.
Whatever.
Well, hash is a direct mapping of data to a memory address, which is a number in the given interval. And most efficient mapping of data to a number is the arithmetic code. So the best hash is made by using the real data statistics in a static model for this mapping. (Well, dynamic model would be perfect, by it'd require remapping of the whole table on every update).
2013-12-28 10:39:56 BennyKulka77 >
I noticed that it's hard to find your site in google, i found it on 19th spot, you should build some quality backlinks to rank it in google and increase traffic. I had the same problem with my website, your should search in google for - k2 seo services - it helped me a lot
2014-11-26 01:06:55 >
2015-01-10 22:18:18 >
|