获取数字的所有约数的最有效方法
可能的重复:
高效查找数字的所有约数
这更多的是一种效率问题而不是通用的“找到一种方法来做到这一点”,但在得到一些奇怪的结果后,我想看看是否有人可以告诉我为什么最后一种方法如此低效:
方法1:暴力,没有优化
public static List<int> proper_divisors(int x)
{
List<int> toreturn = new List<int>();
for (int i = 1; i <= Math.Floor(Math.Sqrt(x)); i++)
{
if (x % i == 0)
{
toreturn.Add(i);
toreturn.Add(x / i);
}
}
if (toreturn.ElementAt(toreturn.Count() / 2) == toreturn.ElementAt(toreturn.Count() / 2 - 1))
{
toreturn.Remove(toreturn.ElementAt(toreturn.Count() / 2));
}
return toreturn;
}
方法2:与之前,但这一次,首先检查它是否是素数(因为这些情况占用了最多的时间,使用 miller-rabin 进行素数检查)
public static List<int> proper_divisors(int x)
{
List<int> toreturn = new List<int>();
if (!isprime(x))
{
for (int i = 1; i <= Math.Floor(Math.Sqrt(x)); i++)
{
if (x % i == 0)
{
toreturn.Add(i);
toreturn.Add(x / i);
}
}
if (toreturn.ElementAt(toreturn.Count() / 2) == toreturn.ElementAt(toreturn.Count() / 2 - 1))
{
toreturn.Remove(toreturn.ElementAt(toreturn.Count() / 2));
}
}
else
{
toreturn.Add(1);
toreturn.Add(x);
}
return toreturn;
}
它认为迄今为止最快的方法是方法 3,因为它减少了它的数量每次找到素数因子时都会进行处理,并且它只尝试素数(这些素数是在运行时由筛子生成的,需要大约 34 毫秒才能获得小于一百万的所有素数),这种方式要做的最后一件事就是采用主要因素及其力量,并列出所有因素。
方式 3:
public static HashSet<int> prime_factors(int x)
{
if (!isprime(x))
{
List<int> toreturn = new List<int>();
int i = 0;
while (primes[i] <= x)
{
if (x % primes[i] == 0)
{
toreturn.Add(primes[i]);
x = x / primes[i];
}
else
{
i++;
}
}
var power_set_primes = GetPowerSet(toreturn);
var factors = new HashSet<int>();
foreach (var p in power_set_primes)
{
var factor = p.Select(z => z).Aggregate(1, (z, y) => z * y);
factors.Add(factor);
}
return factors;
}
else
{
HashSet<int> toreturn = new HashSet<int>();
toreturn.Add(1);
toreturn.Add(x);
return toreturn;
}
public static IEnumerable<IEnumerable<T>> GetPowerSet<T>(List<T> list)
{
return from m in Enumerable.Range(0, 1 << list.Count)
select
from i in Enumerable.Range(0, list.Count)
where (m & (1 << i)) != 0
select list[i];
}
分解前一百万个数字所需的时间: 方式 1:7223 毫秒 方式2:8985 ms(我猜对于小数字来说素数检查不值得) 方式3:49423毫秒
所以我的问题是双重的: 1)为什么方式3这么慢??? 2)有什么东西可以让它更快吗? 顺便说一句,素数被计算为列表,然后转换为数组,因为我认为这样会更快。不好的举动?
Possible Duplicate:
Efficiently finding all divisors of a number
This is much more of an efficiency question than a generic "find a way to do it", but after getting some odd results, I want to see if someone can tell me why the last way is so inefficient:
way 1: brute force, no optimization
public static List<int> proper_divisors(int x)
{
List<int> toreturn = new List<int>();
for (int i = 1; i <= Math.Floor(Math.Sqrt(x)); i++)
{
if (x % i == 0)
{
toreturn.Add(i);
toreturn.Add(x / i);
}
}
if (toreturn.ElementAt(toreturn.Count() / 2) == toreturn.ElementAt(toreturn.Count() / 2 - 1))
{
toreturn.Remove(toreturn.ElementAt(toreturn.Count() / 2));
}
return toreturn;
}
way 2: same as before, but this time, check if its prime first (as those cases take up the most time, using miller-rabin for prime checking)
public static List<int> proper_divisors(int x)
{
List<int> toreturn = new List<int>();
if (!isprime(x))
{
for (int i = 1; i <= Math.Floor(Math.Sqrt(x)); i++)
{
if (x % i == 0)
{
toreturn.Add(i);
toreturn.Add(x / i);
}
}
if (toreturn.ElementAt(toreturn.Count() / 2) == toreturn.ElementAt(toreturn.Count() / 2 - 1))
{
toreturn.Remove(toreturn.ElementAt(toreturn.Count() / 2));
}
}
else
{
toreturn.Add(1);
toreturn.Add(x);
}
return toreturn;
}
what it thought would be the fastest way by far was way 3, because it reduced the number that it was working with every time it found a prime factor, and it only tried primes (these were generated by a sieve at runtime, takes about 34 ms to get all primes less than a million) the last thing this way had to do was take the prime factors and their powers, and make a list of all the factors.
way 3:
public static HashSet<int> prime_factors(int x)
{
if (!isprime(x))
{
List<int> toreturn = new List<int>();
int i = 0;
while (primes[i] <= x)
{
if (x % primes[i] == 0)
{
toreturn.Add(primes[i]);
x = x / primes[i];
}
else
{
i++;
}
}
var power_set_primes = GetPowerSet(toreturn);
var factors = new HashSet<int>();
foreach (var p in power_set_primes)
{
var factor = p.Select(z => z).Aggregate(1, (z, y) => z * y);
factors.Add(factor);
}
return factors;
}
else
{
HashSet<int> toreturn = new HashSet<int>();
toreturn.Add(1);
toreturn.Add(x);
return toreturn;
}
public static IEnumerable<IEnumerable<T>> GetPowerSet<T>(List<T> list)
{
return from m in Enumerable.Range(0, 1 << list.Count)
select
from i in Enumerable.Range(0, list.Count)
where (m & (1 << i)) != 0
select list[i];
}
Time it took to factor the first million numbers:
way 1: 7223 ms
way 2: 8985 ms (prime checking is not worth it for small numbers i guess)
way 3: 49423 ms
so my question is twofold:
1) why is way 3 so slow???
2) is there something that can make it faster?
as an aside, primes was computed as a list, then converted to an array, as I thought it would be faster that way. bad move?
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
这是整数分解的问题域。这里有许多众所周知的算法:
http://en.wikipedia.org/wiki/Integer_factorization# Factoring_algorithms
我建议您选择最佳匹配+配置文件。
我的原始评论:
This is the problem domain of integer factorization. There are a number of wellknown algorithms here:
http://en.wikipedia.org/wiki/Integer_factorization#Factoring_algorithms
I suggest you pick the best match + profile.
my original comment: