(Negative) Log Likelihood Cost Function

Use Cases

This kind of cost function is used mostly when the prediction output is a probability among multiple possibilities.

For example when we are predicting the probability of

• Next character in word
• Object Being a cat, dog, human, ape…
• Call receiving a good/bad rating.

I have seen log likelihood cost function being used with

• Logistic Regression
• Neural Networks for multiclass problems

Negative Likelihood Function For Multiclass Classification