指数拟合与Scipy.Optimise Curve_fit无法正常工作

时间:2014-04-15 11:08:00

标签: python numpy scipy curve-fitting exponential

我试图使用Scipy.Optimise Curve_fit按照简单示例here的方式拟合一些数据。

脚本运行没有错误但是适合性很差。当我在curve_fit的每一步看到popt的输出时,它似乎没有很好地迭代从初始参数跳转到一系列的1.0s,尽管它似乎让第三个参数恢复到一个相当不错的值:< / p>

92.0 0.01 28.0
1.0 1.0 1.0
1.0 1.0 1.0
1.0 1.0 1.0
1.00012207031 1.0 1.0
1.0 1.00012207031 1.0
1.0 1.0 1.00012207031
1.0 1.0 44.3112882656
1.00012207031 1.0 44.3112882656
1.0 1.00012207031 44.3112882656
1.0 1.0 44.3166973584
1.0 1.0 44.3112896048
1.0 1.0 44.3112882656

我不确定是什么导致这种情况,除非我强烈怀疑该模型不适合数据(物理学是物理学)。有什么想法吗?我在下面发布了我的(非常简单的)脚本。感谢。

#!/usr/bin/python

import matplotlib.pyplot as plt
import os
import numpy as np
from scipy.optimize import curve_fit
from matplotlib.ticker import*
from glob import glob
from matplotlib.backends.backend_pdf import PdfPages
import fileinput

path_src=os.getcwd()
dirlist= glob(path_src + '/Gel_Temp_Res.txt')
dirlist.sort()

plots_file='Temp_Curve.pdf'
plots= PdfPages(path_src+'/'+plots_file)

time=[]
temp=[]

for row in fileinput.input(path_src + '/Gel_Temp_Res.txt'):

    time.append(row.split()[0])
    temp.append(row.split()[1])

nptime=np.array(time, dtype='f')
nptemp=np.array(temp, dtype='f')

del time[:]
del temp[:]

# Newton cooling law fitting
def TEMP_FIT(t, T0, k, Troom):
    print T0, k, Troom
    return T0 * np.exp(-k*t) + Troom

y = TEMP_FIT(nptime[41:], nptemp[41]-nptemp[0], 1e-2, nptemp[0])
yn = y + 0.2*np.random.normal(size=len(nptime[41:]))
popt, pcov = curve_fit(TEMP_FIT, nptime[41:], yn)

# Plotting
ax1 = plt.subplot2grid((1,1),(0, 0))
ax1.set_position([0.1,0.1,0.6,0.8])
plt.plot(nptime[41:], nptemp[41:], 'bo--',label='Heater off', alpha=0.5)
plt.plot(nptime[41:], TEMP_FIT(nptime[41:], *popt), label='Newton Cooling Law Fit')
plt.xlim(-25, 250)
plt.xlabel('Time (min)')
plt.ylabel('Temperature ($^\circ$C)')
ax1.grid(True, which='both', axis='both')
plt.legend(numpoints=1, bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
plt.savefig(plots, format='pdf',orientation='landscape')
plt.close()
plots.close()

此外,这是我试图拟合的数据:

100 124
130 120
135 112
140 105
145 99
150 92
155 82
160 75
165 70
170 65
175 60
180 56
185 55
190 52
195 49
200 45
205 44
210 40
215 39
220 37
225 35

1 个答案:

答案 0 :(得分:4)

我清理了你的代码,让时间从0开始,我用指数函数做了不止一次,以使它们正常工作:

import matplotlib.pyplot as plt
import numpy as np
from scipy.optimize import curve_fit

time, temp = np.loadtxt('test.txt', unpack=True)

# Newton cooling law fitting
time -= time.min()
def TEMP_FIT(t, T0, k, Troom):
    print T0, k, Troom
    return T0 * np.exp(-k*t) + Troom

popt, pcov = curve_fit(TEMP_FIT, time, temp)

# Plotting
plt.figure()
plt.plot(time, temp, 'bo--',label='Heater off', alpha=0.5)
plt.plot(time, TEMP_FIT(time, *popt), label='Newton Cooling Law Fit')
plt.xlim(-25, 250)
plt.xlabel('Time (min)')
plt.ylabel('Temperature ($^\circ$C)')
ax = plt.gca()
ax.xaxis.set_ticks_position('bottom')
ax.yaxis.set_ticks_position('left')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.legend(fontsize=8)
plt.savefig('test.png', bbox_inches='tight')

结果是:

enter image description here

删除样本的第一个点:

enter image description here