when we are doing classification problem, ie. the logistic regression, why are we using ‘log’
why are we not doing in the usual way, in which we minimize squared mean?
Log-Livelihood in logistic regression
The answer is related to some hard maths. The loss curve when using Log Loss is convex in shape (with a global minima) while the the curve contains local minimas if MSE is used.
Please refer to this blog. It explains the concept in a much more intuitive way.
Hope this helps!
Hi,
thanks, this blog was very helpful.
1 Like