Week 5 Info: Generative Classifiers
It is now week 5. During this week, we will learn about how most machine learning models can be divided into two distinct classes, discriminative and generative models. Generative models aim to capture the entire data generating process, including that of the covariates, and thus focuses on the probability of the covariates given the class label. Generative models are usually harder to fit than discriminative models and have some other disadvantages, but make up for it through their flexibility, interpretability, and capacity to generate random examples that are like the data (which is often the main application as we have seen for LLMs). We will explore generative models primarily through the example of the Naive Bayes classifier.
You can read more about the plan for the week and the reading/resources in Module 5.
There will be a coding video covering the implementation and interpretation of Naive Bayes.
Lab 3 is due in two weeks, Sunday at midnight.