Logistic regression likelihood estimation(video)

#1

In the “likelihood estimation and loss” video
1.what does the likelihood function mean?
2.How does maximizing the likelihood function gives the optimal set of theta parameters for the classification boundry?

#2

In MLE we try to find that parameter which is most ‘likely’ to be the correct parameter, which gives the maximum value of the likelihood function

For more intuitive explanations i would suggest you reading the two posts below :

https://link.medium.com/vb29Vd8MGT

https://link.medium.com/7wgzxq7NGT

#3

Thanks for the posts I have started to understand likelihood better but I still couldn’t understand:
how maximizing the likelihood function gives the optimal parameters for the boundary line to classify the datasets correctly?

#4

We are using -ve of likelihood as our cost function.
Minimizing our cost function is same as maximizing the likelihood.
And minimizing cost function leads to better decision boundary

#5

In linear regression cost function meant deviation from the data and minimizing the deviation gives the right set of parameters.
But in logistic regression what does the cost function(-ve likelihood) mean it certainly doesn’t seem to be deviation from the true value.

P.s-To me likelihood is some what like probability(please correct me if I am wrong)

#6

The derivation in this article is quite intuitive :

Generally probability and likelihood are used exchangeably