Week 7 Info: Regularization Methods
Week 7 is here and it is time to learn about regularization. Regularization methods were first invented to handle problems in the applied sciences, with first application (that I am aware of) in the geosciences to estiamte the internal structure of the Earth from sparse measurements. Regularization methods have since spread and become highly developed. They allow for models of arbitrary complexity to be used in situations where they would normally overfit. Regularization methods achieve this by penalizing models for assigning values to coefficients different than 0, causing “shrinkage” of coefficients. This increases bias, but reduces variance and thus helps with overfitting. We will learn about the two most important regularization methods, Ridge regression (or an \(L_2\) penalty) and the Lasso (or an \(L_1\) penalty).
You can read more about the plan for the week and the reading/resources in Module 7.
Project Proposal presentations are this week. I have heard from most teams but if I have not heard form you yet make sure to contact me asap.
Lab 4 is due Sunday at midnight.