Python金融大数据分析笔记（自用8）代码加速

一.循环

编写一个函数提取大量随机数，返回平均值，查看执行时间

1.Python

import random
def average_py(n):#定义函数
    s = 0  
    for i in range(n):
        s += random.random()  
    return s / n 
n = 10000000 
%time average_py(n)  
%timeit average_py(n)  
%time sum([random.random() for _ in range(n)]) / n  #使用列表代替函数

2.Numpy

Numpy的优势在于向量化能力

import numpy as np
def average_np(n):
    s = np.random.random(n)  
    return s.mean()
%time average_np(n)
%timeit average_np(n)
s = np.random.random(n)
s.nbytes    

向量化可以加速，但是会占用更多内存

3.Numba

可以动态编译

import numba
average_nb = numba.jit(average_py)  
%time average_nb(n)  
%timeit average_nb(n)  

4.Cython

静态编译

%load_ext Cython
%%cython -a
import random  
def average_cy1(int n):  
    cdef int i  
    cdef float s = 0  
    for i in range(n):
        s += random.random()
    return s / n
%time average_cy1(n)
%timeit average_cy1(n)

二.算法

1.质数

1.1Python

def is_prime(I):
    if I % 2 == 0: return False  
    for i in range(3, int(I ** 0.5) + 1, 2):  
        if I % i == 0: return False  
    return True  
n = int(1e8 + 3)  
n
p1 = int(1e8 + 7)  
p1
%time is_prime(n)
%time is_prime(p1)

1.2Numba

is_prime_nb = numba.jit(is_prime)
%time is_prime_nb(n)  
%time is_prime_nb(p1)

1.3Cython

def is_prime_cy1(I):
    if I % 2 == 0: return False
    for i in range(3, int(I ** 0.5) + 1, 2):
        if I % i == 0: return False
    return True
%timeit is_prime(p1)
%timeit is_prime_cy1(p1)    

使用静态类型声明

def is_prime_cy2(long I):  
    cdef long i  
    if I % 2 == 0: return False
    for i in range(3, int(I ** 0.5) + 1, 2):
        if I % i == 0: return False
    return True
%timeit is_prime_cy2(p1)
%time is_prime_nb(p2)    

多进程

import multiprocessing as mp
pool = mp.Pool(processes=4)  
%time pool.map(is_prime, 10 * [p1])  

2.斐波那契

2.1递归