Python Scipy库入门这一篇就行了

scipy cluster

用于聚类分析，比如K-Means

from numpy import vstack, array
from numpy.random import rand
from scipy.cluster.vq import whiten, kmeans, vq

data = vstack((rand(100, 3) + array([.5, .5, .5]), rand(100, 3)))
# 白化数据
data = whiten(data)

# 指定簇集的个数，获得簇中心
centroids, _ = kmeans(data, 3)

# 根据指定的簇中心对数据进行聚类
clx, _ = vq(data, centroids)
print(clx)

scipy.odr

用于回归分析

import numpy as np
import matplotlib.pyplot as plt
from scipy.odr import *
import random

x = np.array([0, 1, 2, 3, 4, 5])
y = np.array([i ** 2 + random.random() for i in x])

def linear_func(p, x):
    m, c = p
    return m * x + c

linear_model = Model(linear_func)

data = RealData(x, y)

odr = ODR(data, linear_model, beta0=[0., 1.])

out = odr.run()
out.pprint()
beta = out.beta
x_track = np.linspace(-1, 6, 100)
y_track = beta[0] * x_track + beta[1]
plt.plot(x, y, 'r*', x_track, y_track, 'b-')
plt.show()

scipy.stats

用于统计分析

生成服从指定分布的随机数

stats.norm.rvs(loc ,scale ,size )

loc：正态分布的期望

scale：正态分布的标准差

size：指定随机数的形状参数

# 生成服从于正态分布的随机数
import numpy as np
import scipy.stats as st

random_array = st.norm.rvs(loc=0, scale=0.1, size=(5, 5))
print(random_array)

'''
[[ 0.08032507  0.04995535 -0.11339123  0.04988461 -0.00184663]
 [-0.06463444 -0.16271425 -0.05429032 -0.05717832  0.14683614]
 [ 0.07611145 -0.07284285 -0.06833373 -0.05064769  0.16348252]
 [ 0.0632136   0.25100341 -0.04703659 -0.11957341  0.00846529]
 [ 0.050868    0.10585317  0.01720439  0.07208371 -0.00230902]]
'''

求概率密度函数指定点的函数值

stats.norm.pdf(point, loc, scale)

point: 点

loc: 正态分布的标准差

scale:正态分布的期望

import numpy as np
import scipy.stats as st

norm_res = st.norm.pdf([0, 1, 2], loc=0, scale=1)
print(norm_res)

'''
[0.39894228 0.24197072 0.05399097]
'''

求累计分布函数指定点的函数值

stats.norm.cdf(point, loc, scale)

point:点

loc: 正态分布的标准差

scale:正态分布的期望

import numpy as np
import scipy.stats as st

res = st.norm.cdf(0, loc=0, scale=1)
print(res)
'''
0.5
'''

scipy.special

提供数学函数，这里给出不常见的函数

函数名	描述
comb(N,M)	组合数
perm(N,M)	排列数
beta(a,b)	beta函数

scipy.linalg

用于矩阵求解运算

下面求解本线性方程组
$$\begin{gathered} x1+4\ast x2+2\ast x3=7 \\ 4\ast x1+2\ast x2+6\ast x3=23 \\ 8\ast x1+23\ast x2+5\ast x3=5 \end{gathered}$$

from scipy import linalg as lg
import numpy as np

mat = [[1, 4, 2], [4, 2, 6], [8, 23, 5]]
res = [7, 23, 5]
# 求矩阵的行列式
print(lg.det(mat))
'''
136.0
'''
# 求矩阵的特征值
print(lg.eig(mat))
'''
(array([17.29893809+0.j, -0.94058671+0.j, -8.35835137+0.j]), array([[ 0.20841245,  0.85648921,  0.034842  ],
       [ 0.40383822, -0.17223167, -0.51096382],
       [ 0.89077435, -0.48658245,  0.85889581]]))
'''
# 求矩阵的逆矩阵
print(lg.inv(mat))
'''
[[-0.94117647  0.19117647  0.14705882]
 [ 0.20588235 -0.08088235  0.01470588]
 [ 0.55882353  0.06617647 -0.10294118]]
'''
# 求解线性方程组
print(lg.solve(mat, res))
'''
[-1.45588235 -0.34558824  4.91911765]
'''
# 利用Numpy库求矩阵的秩
print(np.linalg.matrix_rank(mat))
'''
3
'''

scipy.interpolate

用于插值计算

fun = interp1d(x, y, kind)

x: 横坐标

y:纵坐标

kind:差值方法

kind	描述
nearest	最邻近插值法
zero	阶梯插值
slinear、linear	线性插值
quadratic、cubic	2、3阶B样条曲线插值

from scipy.interpolate import interp1d
import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(1, 20, 12)

y = np.cos(x)

f1 = interp1d(x, y, kind='cubic')
f2 = interp1d(x, y, kind='nearest')
f3 = interp1d(x, y, kind='linear')
f4 = interp1d(x, y, kind='quadratic')

x_index = np.linspace(1, 20, 200)
y_index = f1(x_index)
plt.subplot(2, 2, 1)
plt.title('cubic')
plt.plot(x, y, 'g*')
plt.plot(x_index, y_index, 'b--')

x_index = np.linspace(1, 20, 200)
y_index = f2(x_index)
plt.subplot(2, 2, 2)
plt.title('nearest')
plt.plot(x, y, 'g*')
plt.plot(x_index, y_index, 'r--')

x_index = np.linspace(1, 20, 200)
y_index = f3(x_index)
plt.subplot(2, 2, 3)
plt.title('linear')
plt.plot(x, y, 'g*')
plt.plot(x_index, y_index, 'm--')

x_index = np.linspace(1, 20, 200)
y_index = f4(x_index)
plt.subplot(2, 2, 4)
plt.title('quadratic')
plt.plot(x, y, 'g*')
plt.plot(x_index, y_index, 'k--')

plt.show()

多元插值以及更多请参考这篇文章

scipy.integrate

用于计算多重积分

import numpy as np
from scipy.integrate import tplquad, dblquad, quad
# 返回结果和误差
val1, err1 = quad(lambda x: np.sin(x), 0, np.pi)  # x上界pi
print('积分结果：', val1)
'''
积分结果： 2.0
'''
# 二重积分
val2, err2 = dblquad(lambda x, y: np.sin(x) * np.cos(y / 2), 0, 2, 4, 6)
print(val2)
'''
-2.7159551561049335
'''
# 三重积分
val3, err3 = tplquad(lambda z, y, x: np.sin(x) / np.sqrt(x + y ** 2) + np.cos(z ** 3), -3, 5, 2, 4, 3, 4)
print(val3)
'''
-1.2868508936407685
'''