Python求离散序列导数的示例

from scipy.optimize import leastsq
class search(object):
  def __init__(self, filename):
    self.filename = filename

  def func(self, x, p):
    f = np.poly1d(p)
    return f(x)

  def residuals(self, p, x, y, reg):
    regularization = 0.1 # 正则化系数lambda
    ret = y - self.func(x, p)
    if reg == 1:
      ret = np.append(ret, np.sqrt(regularization) * p)
    return ret

  def LeastSquare(self, data, k=100, order=4, reg=1, show=1): # k为求导窗口宽度,order为多项式阶数,reg为是否正则化
    l = self.len
    step = 2 * k + 1
    p = [1] * order
    for i in range(0, l, step):
      if i + step < l:
        y = data[i:i + step]
        x = np.arange(i, i + step)
      else:
        y = data[i:]
        x = np.arange(i, l)
      try: 
        r = leastsq(self.residuals, p, args=(x, y, reg))
      except:
        print("Error - curve_fit failed")
      fun = np.poly1d(r[0]) # 返回拟合方程系数
      df_1 = np.poly1d.deriv(fun) # 求得导函数
      df_2 = np.poly1d.deriv(df_1)
      df_3 = np.poly1d.deriv(df_2)
      df_value = df_1(x)
      df3_value = df_3(x)

  def gaussian(self, x, *param):
    fun = param[0]*np.exp(-np.power(x - param[2], 2.) / (2 * np.power(param[4],    2.)))+param[1]*np.exp(-np.power(x - param[3], 2.) / (2 * np.power(param[5], 2.)))
    return fun