Error function :

#1

Hello sir,
Why did we take the negative of the log likelihood as the error function?

#2

Hey Abishek,
As discussed in the video we have log likelihood, which we have to maximize to get the optimal theta values. But in Machine Learning we always find a cost function that we minimize, in logistic regression case since we want to maximize the log likelihood, can we say that take the negative of log likelihood, treat it like a cost function and thus minimize it using gradient descent.
So basically to form a cost function we put negative sign before the log likelihood.

I hope this clears your doubts
Thanks :slight_smile:

1 Like
#3

Hello Sir,
Thanks for the response…I got it!