On 30/08/2015 00:18, "Claude Warren" <[email protected]> wrote:

>Well to start with create one bloom filter for each quad.  Using
>http://hur.st/bloomfilter?n=4&p=1.0E-6 you can see that 4 items with a 1
>on
>1 million collision rate requires 15 bytes.  So each bloom filter is
>15bytes.

n is the number of distinct items you expect the filter to see so it is
very unlikely to be 4

Once you have added n items the false positive rate will start to go
upwards

You probably want to pick a more sensible n like 100,000
(http://hur.st/bloomfilter?n=10000&p=1.0E-6) which would require ~35
kilobytes

Rob




Reply via email to