加快迭代范围并比较值

时间:2018-09-25 14:57:06

标签: python python-3.x sum comparison list-comprehension

我有一个循环,该循环生成数字组合,并求解以查看它们是否等于解。运行探查器后,我发现求和与解决方案的比较花费了最长的时间。这是仅是由于每个呼叫的数量,还是有一种方法可以使其更快?

Profiler输出: [名称,呼叫计数,时间(ms),自己的时间(ms)]

[listcomp,23114767,5888,5888]占总时间的25%。

[内置总和,23164699、3097、3097]占总时间的12%

I've attached an image of the profiler

现在我想不出一种缩小搜索范围的方法,因此试图节省其他时间:)

感谢您的帮助。

rangedict = {'f1': range(1850, 1910), 'f2': range(2401, 2482), 'f3': range(5150, 5850)}
coefficient = [-3, 1, 1]
possiblesolution = 1930

for combination in itertools.product(*rangedict.values()):
    if solutionfound:
        break
    else:
        currentsolution = sum([x*y for x, y in zip(coefficient, combination)])
        if currentsolution == possiblesolution:
            solutionfound = True
            dosomething()

2 个答案:

答案 0 :(得分:0)

如评论中所述:

:将系数直接打包到ranges中将使整个过程加快一点:

from itertools import product

possiblesolution = 1930
solutionfound = False
rangedict2 = {'f1': range(-3*1850, -3*1910, -3),
              'f2': range(2401, 2482), 
              'f3': range(5150, 5850)}

for combination in product(*rangedict2.values()):
    if sum(combination) == possiblesolution:
        solutionfound = True
        print(combination[0]//(-3), combination[1], combination[2])
        break

或完全不同的方法:创建一个f1f2可获得的总和的字典,然后检查是否可以达到目标possiblesolution

from collections import defaultdict

rangedict3 = {'f1': range(-3*1850, -3*1910, -3), 
              'f2': range(2401, 2482), 
              'f3': range(5150, 5850)}

sums = {item: [[item]] for item in rangedict3['f1']}
# sums = {-5550: [[-5550]], -5553: [[-5553]], ...}

new_sums = defaultdict(list)
for sm, ways in sums.items():
    for item in rangedict3['f2']:
        new_sum = sm + item
        for way in ways:
           new_sums[new_sum].append(way + [item])
# new_sums = {-3149: [[-5550, 2401], [-5553, 2404], ...],
#             -3148: [[-5550, 2402], [-5553, 2405], ...],
#             ....}         

for item in rangedict3['f3']:
    if possiblesolution - item in new_sums:
        f1_f2 = new_sums[possiblesolution - item]
        print(f1_f2[0][0]//(-3), f1_f2[0][1], item)
        # print(new_sums[possiblesolution - item], item)
        break

那样,您还可以轻松获得其余解决方案。


或仅将f2f3一起:

f1 = rangedict3['f1']
f2 = rangedict3['f2']
f3 = rangedict3['f3']

# the sums that are reachable from f2 and f3
f2_f3 = range(2401 + 5150, 2482 + 5850)

for item in f1:
    if possiblesolution - item in f2_f3:
        pmi = possiblesolution - item
        x1 = item//(-3)
        for x2 in f2:
            if pmi-x2 in f3:
                x3 = pmi-x2
                break
        print(x1, x2, x3)
        break

和最后一个次要的提速:如果您真的只需要一个解决方案,则x2x3仅有4种(甚至3种)可能的情况:

f1 = rangedict3['f1']
f2 = rangedict3['f2']
f3 = rangedict3['f3']

min_x2 = min(f2)
max_x2 = max(f2)
min_x3 = min(f3)
max_x3 = max(f3)

# the sums that are reachable from f2 and f3
f2_f3 = range(2401 + 5150, 2482 + 5850 - 1)

for item in f1:
    if possiblesolution - item in f2_f3:
        pmi = possiblesolution - item
        x1 = item//(-3)

        if pmi-min_x2 in f3:
            x2 = min_x2
            x3 = pmi-x2
        elif pmi-max_x2 in f3:
            x2 = max_x2
            x3 = pmi-x2
        elif pmi-min_x3 in f2:
            x3 = min_x3
            x2 = pmi-x3
        # elif pmi-max_x3 in f2:
        else:
            x3 = max_x3
            x2 = pmi-x3

        print(x1, x2, x3)
        break

答案 1 :(得分:0)

  1. 您正在为每个组合重做乘法。最好做

    bound_dict =  [['f1', [1850, 1910]], ['f2', [2401, 2482]], ['f3', [5150, 5850]]
    coefficient = [-3, 1, 1]
    rangedict2={bound_dict(i)[0]:range(coefficient[i]*bound_dict(i)[1][0],coefficient[i]*bound_dict(i)[1][0],coefficient[i]) for i in range(len(coefficient))}
    
  2. 您正在遍历所有组合,而忽略了约束消除了自由度。您应删除元素数量最多的范围,在本例中为第三个元素。这样会将迭代次数减少700倍。

    for combination in itertools.product(*rangedict2.values()[:-1]):
        currentsolution = sum(combination)
        if  possiblesolution-currentsolution in rangedict2.values()[:-1]:
            solutionfound = True
            dosomething()
            break     
    
  3. 您可以尝试进行列表理解:

    [possiblesolution-currentsolution in rangedict2.values()[:-1] for #etc.
    
  4. 如果您执行的数字多于三个,则可以尝试进行嵌套的for循环,而不要使用combination。一般的实现是这样的:

    last_value1 = list(rangedict2.values()[0])[0]       
    last_value2 = list(rangedict2.values()[1])[0] 
    current_sum = last_value1+last_value2
    for value1 in rangedict2.values()[0]:
        current_sum = current_sum+value1-last_value1
        for value2 in rangedict2.values()[1]:
            current_sum = current_sum +value2-last_value2
            #check whether this is a solution
            last_value2 = value2
            #more nested for-loops
        last_value1 = value1
    

这样,无论rangedict2有多长,每个迭代仅执行两个加法运算,而不是len(rangedict2)-1运算。在这种情况下,两种情况下的结果均为2,但是在一般情况下,这可能会减少操作次数。在这种情况下,由于差异始终相同,因此您可以执行以下操作:

    for value1 in (rangedict2.values()[0])):
        for current_sum in range(value1+2401,value2+2482):
            #check whether this is a solution
  1. 看看您的实际数字,您只需做一些数学运算就能简化组合的数量。例如,从最后两个范围获得的唯一总和是2401 + 5150至2482 + 5850-2。您不需要为此遍历81 * 700个组合;您只需要81 + 700。