python 数学工具包括:
1.函数的逼近
1.1.回归
1.2.插值
2.凸优化
3.积分
4.符号数学
本文介绍函数的逼近的回归方法
1.作为基函数的单项式
对函数
的拟合
首先定义函数并且可视化
import numpy as np import matplotlib.pyplot as plt import pandas as pd def f(x): return np.sin(x)+0.5*x x=np.linspace(-2*np.pi,2*np.pi,50) plt.plot(x,f(x),'b') plt.xlabel('x') plt.ylabel('f(x)') plt.grid(True)
先用一次函数拟合
reg=np.polyfit(x,f(x),deg=1) ry=np.polyval(reg,x) plt.plot(x,f(x),'b',label='f(x)') plt.grid(True) plt.plot(x,ry,'r.',label='reg') plt.legend(loc=0)
再用高次函数进行拟合
reg=np.polyfit(x,f(x),deg=16) ry=np.polyval(reg,x) plt.plot(x,f(x),'b',label='f(x)') plt.grid(True) plt.plot(x,ry,'r.',label='reg') plt.legend(loc=0)
拟合效果的检查
print('平均误差:',sum((ry-f(x))**2)/len(x)) 平均误差: 3.16518401761e-13
np.allclose(ry,f(x)) True
2.单独的基函数
首先常见一个空的矩阵,然后为任一行添加函数
mat=np.zeros((3+1,len(x))) mat[3,:]=x**3 mat[2,:]=x**2 mat[1,:]=x mat[0,:]=1
reg=np.linalg.lstsq(mat.T,f(x))
#输出系数
reg[0]
array([ 1.52685368e-14, 5.62777448e-01, -1.11022302e-15, -5.43553615e-03])
#输出图形 ry=np.dot(reg[0],mat) plt.plot(x,f(x),'b',label='f(x)') plt.plot(x,ry,'r.',label='reg') plt.grid(True) plt.legend(loc=0)
对每行的基函数进行变换:
mat=np.zeros((3+1,len(x))) mat[3,:]=np.sin(x) mat[2,:]=x**2 mat[1,:]=x mat[0,:]=1 reg=np.linalg.lstsq(mat.T,f(x)) ry=np.dot(reg[0],mat) plt.plot(x,f(x),'b',label='f(x)') plt.plot(x,ry,'r.',label='reg') plt.grid(True) plt.legend(loc=0)
3.多维情形
def fm(x,y): return np.sin(x) + 0.25 * x + np.sqrt(y) + 0.05**y*2 x = np.linspace(0, 10, 20) y = np.linspace(0, 10, 20) x, y = np. meshgrid( x, y) Z = fm(x,y) x = x.flatten() y = x. flatten() import statsmodels.api as sm matrix=np.zeros((len(x),6+1)) matrix[:,6] = np.sqrt(y) matrix[:,5] = np.sin(x) matrix[:,4] = y**2 matrix[:,3] = y**2 matrix[:,2] = y matrix[:,1] = x matrix[:,0] = 1 res=sm.OLS(fm(x,y),matrix).fit() print(res.summary().as_text())
OLS Regression Results ============================================================================== Dep. Variable: y R-squared: 0.999 Model: OLS Adj. R-squared: 0.999 Method: Least Squares F-statistic: 9.605e+04 Date: Tue, 31 Jul 2018 Prob (F-statistic): 0.00 Time: 10:51:36 Log-Likelihood: 661.47 No. Observations: 400 AIC: -1313. Df Residuals: 395 BIC: -1293. Df Model: 4 Covariance Type: nonrobust ============================================================================== coef std err t P>|t| [0.025 0.975] ------------------------------------------------------------------------------ const 1.9548 0.010 193.732 0.000 1.935 1.975 x1 0.5891 0.005 111.546 0.000 0.579 0.600 x2 0.5891 0.005 111.546 0.000 0.579 0.600 x3 -0.0150 0.000 -54.014 0.000 -0.016 -0.014 x4 -0.0150 0.000 -54.014 0.000 -0.016 -0.014 x5 0.9533 0.004 251.168 0.000 0.946 0.961 x6 -1.6190 0.020 -79.979 0.000 -1.659 -1.579 ============================================================================== Omnibus: 4.352 Durbin-Watson: 0.880 Prob(Omnibus): 0.113 Jarque-Bera (JB): 4.214 Skew: -0.208 Prob(JB): 0.122 Kurtosis: 2.717 Cond. No. 4.93e+17 ==============================================================================