蓄水池算法简介

 
 

  
  从N个元素中随机抽取k个元素，但是N不定。每个元素抽取的概率是k/N。
 
 
 
 


 
 
 
 

  
  解答：先选中前k个，从k+1到最后一个元素，以1/i (i = k+1, k+2, ... N)的概率选中第i个元素，并且随机替换一个原来选中的元素。
 
 

'''
 Reservoir Sampling implementation
'''
from random import Random

def RandomSelect(knum, rand=None):
    ''' (int, func) -> list

    Reservoir Sampling implementation
    >>> myList = [1,2,3,4,5,6,7,8,-1] 
    >>> ge = RandomSelect(3)
    >>> ge.next()
    >>> for v in myList:
            cr.send(v)
    '''
    selection = None
    k_elems_list = []
    count = 0
    if rand is None:
        rand = Random()
    while True:
        item = yield selection
        if item == -1:
            break
        if len(k_elems_list) < knum:
            k_elems_list.append(item)
        elif rand.randint(0, count) == 0:
            index = rand.randint(0, knum-1)
            k_elems_list.pop(index)
            k_elems_list.insert(index, item)
        count += 1
    print k_elems_list

if __name__ == '__main__':
    myList = [1,2,3,4,5,6,7,8,-1]
    cr = RandomSelect(3);
    cr.next() # advance to the yield statement, otherwise I can't call send
    try:
        for val in myList:
            cr.send(val)
    except StopIteration:
        pass
    finally:
        del cr