Recently used the scipy least_square method of polynomial fitting, feel with the old version of the original scipy. Optimize. Compared leastsq, this version of the method, the initial value of the setting for a great influence on the result, the same initial value setting, the fitting results are not identical,

Baidu to check the case of most of the least-square method used is the old version leastsq, do not know to have people used the new least_square? What's the matters needing attention when use?