Loss Functions

Question 1

What are the two types of loss functions?

Answer 1

Probabilistic Loss Functions

Non-Probabilistic Loss Functions

Question 2

Why is a Regularisation term added to the Linear Least Squares approach?

Answer 2

To avoid Over Fitting

Question 3

What is Likelihood?

Answer 3

It is the probability of the data, given the parameters

P( data | parameters )

A model can be trained by maximising the likelihood, i.e finding the parameters that make the data most likely

Question 4

When using Maximum Likelihood Estimation , what is one assumption we often make about the distribution of the data?

Answer 4

We often assume the data follows a Gaussian distribution, or a Bernoulli Distribution for binary classification, or a Multinomial distribution for multi-class classification

Then we just have to find the variance and mean that make the data most likely