What is the mathematical principal of setting class weight in logestic regression in scikit-learn library in Python

[ScienceCowboy]

Dear all, 

In machine learning library scikit-learn of python, the logestic regression function has an argument "class_weight". When you set a higher class weight to a class during fitting the logestic model, you will get higher predictive accuracy of this class.  I wish to know what is mathematical principal of setting class_weight. Is it related to modify the target function of logestic regression (https://drive.google.com/open?id=16TKZFCwkMXRKx_fMnn3d1rvBWwsLbgAU) ?

Thank you in advance.