我需要实现一个布隆过滤器。我不能找到出路了这一点。

I need to implement a bloom filter. And I cannot find a way out of this.

随着功能的固定电话号码，我怎么能计算布鲁姆的大小过滤器给误报的概率是多少？

With fixed number of functions, how can I calculate size of a Bloom Filter given the probability of false positives ?

例如，我想，该过滤器有误报10％，我有一些功能，集合中元素的个数。

For example, I want that the filter have 10% of false positives, I have the number functions and the number of elements in the set.

我如何计算布隆过滤器相匹配的假阳性概率的大小？

How can I calculate the size of Bloom Filter that match the false positive probability ?

推荐答案

是在维基百科。假设你有足够的哈希函数可用，需要〜每件4.8位上给出的0.1指定的假阳性率。

The formula for this is on the Wikipedia. Assuming you have enough hash functions available, you need ~4.8 bits per element given the false positive rate you specified of 0.1.

在这种情况下，它看起来像4散列函数将是最佳的。需要注意的是更多的hash函数并不总是更好 - 如果有相对滤波器的尺寸非常多的散列函数，可以快速设置几乎所有的位上，你会得到很多误报

In this case it looks like 4 hash functions would be optimal. Note that more hash functions isn't always better -- if there are very many hash functions relative to the size of the filter, you quickly set almost all the bits on, and you get lots of false positives.