Dr. Shamsiiat (Shamsi) Abdurakhmanova discusses linear regression methods, which are among the most widely used machine learning methods. Linear regression methods learn a linear hypothesis map whose predictions incur minimum (average) loss on a given training set of labeled data point. We obtain different linear regression methods by using different loss functions.
The choice of loss function in regression methods crucially influences their statistical and computational properties. As a point in case, using the average squared error loss ("MSE") is computationally appealing as it allows for implementations with low computational complexity. On the other hand, using the (computationally less appealing) average absolute error loss ("MAE") is statistically appealing as it results in ML methods that are robust against the presence of outliers in the training set.
Read more about linear regression methods in Chapter 3 of
A. Jung, "Machine Learning: The Basics,", Springer, Singapore, 2022. (draft: mlbook.cs.aalto.fi)