【发布时间】：2020-07-27 22:15:11
【问题描述】：
给定一个整数 n，以及 2 个实数序列 {a_1, ..., a_n} 和 {b_1 em>, ..., b_n}，对于所有 i，a_i, b_i > 0。对于给定的固定 m n 让 {P_1, ..., P_m} 是将 {1, ..., n} 设置为 P_1 U ... U P_n = {1, ..., n}，与 P_i 的成对不相交（空相交）。我希望找到一个大小为 m 的分区来最大化表达式
集合的分区数是n选择m，蛮力做的大得令人望而却步。有没有更好的迭代或近似解决方案？
为了深入了解这个问题，最后的代码块通过蛮力解决。对于实际的尺寸问题（n ~ 1e6, k ~ 20），它无法按原样使用，但很容易分散。
编辑：按a^2/ba、b进行预排序/em> 总是给出增加的分区索引：
a = rng.uniform(low=0.0, high=10.0, size=NUM_POINTS)
b = rng.uniform(low=0.0, high=10.0, size=NUM_POINTS)
ind = np.argsort(a/b)
(a,b) = (seq[ind] for seq in (a,b))
使用
运行的示例NUM_POINTS = 16
PARTITION_SIZE = 3
给出一个最优的分区
[[0, 1, 2, 3, 4, 5, 6, 7], [8, 9], [10, 11]]
在索引中是单调的。我想我可以证明这一点。如果是这样，蛮力搜索可以改进为 n 选择 k-1 次，仍然很长，但可以节省大量资金。
import numpy as np
import multiprocessing
import concurrent.futures
from functools import partial
from itertools import islice
rng = np.random.RandomState(55)
def knuth_partition(ns, m):
def visit(n, a):
ps = [[] for i in range(m)]
for j in range(n):
ps[a[j + 1]].append(ns[j])
return ps
def f(mu, nu, sigma, n, a):
if mu == 2:
yield visit(n, a)
else:
for v in f(mu - 1, nu - 1, (mu + sigma) % 2, n, a):
yield v
if nu == mu + 1:
a[mu] = mu - 1
yield visit(n, a)
while a[nu] > 0:
a[nu] = a[nu] - 1
yield visit(n, a)
elif nu > mu + 1:
if (mu + sigma) % 2 == 1:
a[nu - 1] = mu - 1
else:
a[mu] = mu - 1
if (a[nu] + sigma) % 2 == 1:
for v in b(mu, nu - 1, 0, n, a):
yield v
else:
for v in f(mu, nu - 1, 0, n, a):
yield v
while a[nu] > 0:
a[nu] = a[nu] - 1
if (a[nu] + sigma) % 2 == 1:
for v in b(mu, nu - 1, 0, n, a):
yield v
else:
for v in f(mu, nu - 1, 0, n, a):
yield v
def b(mu, nu, sigma, n, a):
if nu == mu + 1:
while a[nu] < mu - 1:
yield visit(n, a)
a[nu] = a[nu] + 1
yield visit(n, a)
a[mu] = 0
elif nu > mu + 1:
if (a[nu] + sigma) % 2 == 1:
for v in f(mu, nu - 1, 0, n, a):
yield v
else:
for v in b(mu, nu - 1, 0, n, a):
yield v
while a[nu] < mu - 1:
a[nu] = a[nu] + 1
if (a[nu] + sigma) % 2 == 1:
for v in f(mu, nu - 1, 0, n, a):
yield v
else:
for v in b(mu, nu - 1, 0, n, a):
yield v
if (mu + sigma) % 2 == 1:
a[nu - 1] = 0
else:
a[mu] = 0
if mu == 2:
yield visit(n, a)
else:
for v in b(mu - 1, nu - 1, (mu + sigma) % 2, n, a):
yield v
n = len(ns)
a = [0] * (n + 1)
for j in range(1, m + 1):
a[n - m + j] = j - 1
return f(m, n, 0, n, a)
def Bell_n_k(n, k):
''' Number of partitions of {1,...,n} into
k subsets, a restricted Bell number
'''
if (n == 0 or k == 0 or k > n):
return 0
if (k == 1 or k == n):
return 1
return (k * Bell_n_k(n - 1, k) +
Bell_n_k(n - 1, k - 1))
NUM_POINTS = 13
PARTITION_SIZE = 4
NUM_WORKERS = multiprocessing.cpu_count()
INT_LIST= range(0, NUM_POINTS)
REPORT_EACH = 10000
partitions = knuth_partition(INT_LIST, PARTITION_SIZE)
# Theoretical number of partitions, for accurate
# division of labor
num_partitions = Bell_n_k(NUM_POINTS, PARTITION_SIZE)
bin_ends = list(range(0,num_partitions,int(num_partitions/NUM_WORKERS)))
bin_ends = bin_ends + [num_partitions] if num_partitions/NUM_WORKERS else bin_ends
islice_on = list(zip(bin_ends[:-1], bin_ends[1:]))
# Have to consume it; can't split work on generator
partitions = list(partitions)
rng.shuffle(partitions)
slices = [list(islice(partitions, *ind)) for ind in islice_on]
return_values = [None] * len(slices)
futures = [None] * len(slices)
a = rng.uniform(low=0.0, high=10.0, size=NUM_POINTS)
b = rng.uniform(low=0.0, high=10.0, size=NUM_POINTS)
ind = np.argsort(a/b)
(a,b) = (seq[ind] for seq in (a,b))
def start_task():
print('Starting ', multiprocessing.current_process().name)
def _task(a, b, partitions, report_each=REPORT_EACH):
max_sum = float('-inf')
arg_max = -1
for ind,part in enumerate(partitions):
val = 0
for p in part:
val += sum(a[p])**2/sum(b[p])
if val > max_sum:
max_sum = val
arg_max = part
if not ind%report_each:
print('Percent complete: {:.{prec}f}'.
format(100*len(slices)*ind/num_partitions, prec=2))
return (max_sum, arg_max)
def reduce(return_values):
return max(return_values, key=lambda x: x[0])
task = partial(_task, a, b)
with concurrent.futures.ThreadPoolExecutor() as executor:
for ind,slice in enumerate(slices):
futures[ind] = executor.submit(task, slice)
return_values[ind] = futures[ind].result()
reduce(return_values)
【问题讨论】：
-
谢谢 - 编辑了问题。
-
你能给一些预期的输入/输出示例吗？
-
添加了蛮力解决方案以进一步了解行为。
标签： algorithm partition approximation