counting combinations and permutations efficiently

if n is not far from r then using the recursive definition of combination is probably better, since xC0 == 1 you will only have a few iterations:

The relevant recursive definition here is:

nCr = (n-1)C(r-1) * n/r

This can be nicely computed using tail recursion with the following list:

[(n - r, 0), (n - r + 1, 1), (n - r + 2, 2), ..., (n - 1, r - 1), (n, r)]

which is of course easily generated in Python (we omit the first entry since nC0 = 1) by izip(xrange(n - r + 1, n+1), xrange(1, r+1)) Note that this assumes r <= n you need to check for that and swap them if they are not. Also to optimize use if r < n/2 then r = n - r.

Now we simply need to apply the recursion step using tail recursion with reduce. We start with 1 since nC0 is 1 and then multiply the current value with the next entry from the list as below.

from itertools import izip

reduce(lambda x, y: x * y[0] / y[1], izip(xrange(n - r + 1, n+1), xrange(1, r+1)), 1)

Two fairly simple suggestions:

1. To avoid overflow, do everything in log space. Use the fact that log(a * b) = log(a) + log(b), and log(a / b) = log(a) - log(b). This makes it easy to work with very large factorials: log(n! / m!) = log(n!) - log(m!), etc.

2. Use the gamma function instead of factorial. You can find one in scipy.stats.loggamma. It's a much more efficient way to calculate log-factorials than direct summation. loggamma(n) == log(factorial(n - 1)), and similarly, gamma(n) == factorial(n - 1).

If you don't need a pure-python solution, gmpy2 might help (gmpy2.comb is very fast).

For Python until 3.7:

def prod(items, start=1):
    for item in items:
        start *= item
    return start


def perm(n, k):
    if not 0 <= k <= n:
        raise ValueError(
            'Values must be non-negative and n >= k in perm(n, k)')
    else:
        return prod(range(n - k + 1, n + 1))


def comb(n, k):
    if not 0 <= k <= n:
        raise ValueError(
            'Values must be non-negative and n >= k in comb(n, k)')
    else:
        k = k if k < n - k else n - k
        return prod(range(n - k + 1, n + 1)) // math.factorial(k)

For Python 3.8+:

• math.perm()
• math.comb()

Interestingly enough, some manual implementation of the combination function may be faster than math.comb():

def math_comb(n, k):
    return math.comb(n, k)


def comb_perm(n, k):
    k = k if k < n - k else n - k
    return math.perm(n, k) // math.factorial(k)


def comb(n, k):
    k = k if k < n - k else n - k
    return prod(range(n - k + 1, n + 1)) // math.factorial(k)


def comb_other(n, k):
    k = k if k > n - k else n - k
    return prod(range(n - k + 1, n + 1)) // math.factorial(k)


def comb_reduce(n, k):
    k = k if k < n - k else n - k
    return functools.reduce(
        lambda x, y: x * y[0] // y[1],
        zip(range(n - k + 1, n + 1), range(1, k + 1)),
        1)


def comb_iter(n, k):
    k = k if k < n - k else n - k
    result = 1
    for i in range(1, k + 1):
        result = result * (n - i + 1) // i
    return result


def comb_iterdiv(n, k):
    k = k if k < n - k else n - k
    result = divider = 1
    for i in range(1, k + 1):
        result *= (n - i + 1)
        divider *= i
    return result // divider


def comb_fact(n, k):
    k = k if k < n - k else n - k
    return math.factorial(n) // math.factorial(n - k) // math.factorial(k)

so that actually comb_perm() (implemented with math.perm() and math.factorial()) is actually faster than math.comb() most of the times for these benchamarks, which show the computation time for fixed n=256 and increasing k (up until k = n // 2).

Note that comb_reduce(), which is quite slow, is essentially the same approach as from @wich's answer, while comb_iter(), also relatively slow, is essentially the same approach as @ZXX's answer.

Partial analysis here (without comb_math() and comb_perm() since they are not supported in Python's version of Colab -- 3.7 -- as of last edit).

If you are computing N choose K (which is what I think you're doing with ncr), there is a dynamic programming solution that may be a lot faster. This will avoid factorial, plus you can keep the table if you want for later use.

Here is a teaching link for it:

http://www.csc.liv.ac.uk/~ped/teachadmin/algor/dyprog.html

I am unsure of how to better solve your first problem, though, sorry.

Edit: Here is the mock-up. There are some pretty hilarious off-by-one errors, so it can certainly stand some more clean up.

import sys
n = int(sys.argv[1])+2#100
k = int(sys.argv[2])+1#20
table = [[0]*(n+2)]*(n+2)

for i in range(1,n):
    table[i][i] = 1
for i in range(1,n):
    for j in range(1,n-i):
        x = i+j
        if j == 1: table[x][j] = 1
        else: table[x][j] = table[x-1][j-1] + table[x-1][j]

print table[n][k]

If your problem does not require knowing the exact number of permutations or combinations, then you could use Stirling's approximation for the factorial.

That would lead to code like this:

import math

def stirling(n):
    # http://en.wikipedia.org/wiki/Stirling%27s_approximation
    return math.sqrt(2*math.pi*n)*(n/math.e)**n

def npr(n,r):
    return (stirling(n)/stirling(n-r) if n>20 else
            math.factorial(n)/math.factorial(n-r))

def ncr(n,r):    
    return (stirling(n)/stirling(r)/stirling(n-r) if n>20 else
            math.factorial(n)/math.factorial(r)/math.factorial(n-r))

print(npr(3,2))
# 6
print(npr(100,20))
# 1.30426670868e+39
print(ncr(3,2))
# 3
print(ncr(100,20))
# 5.38333246453e+20

from numpy import prod

def nCr(n,r):
    numerator = range(n, max(n-r,r),-1)
    denominator = range(1, min(n-r,r) +1,1)
    return int(prod(numerator)/prod(denominator))

Using xrange() instead of range() will speed things up slightly due to the fact that no intermediate list is created, populated, iterated through, and then destroyed. Also, reduce() with operator.mul.

For N choose K you could use Pascals triangle. Basically you would need to keep array of size N around to compute all the N choose K values. Only additions would be required.