Machine Learning

machine learning

develops algorithms designed to be applied to datasets, focusing on prediction, classification, and clustering or grouping tasks

unsupervised learning

goal is to divide high-dimensional data into clusters that are similar in their set of features

unsupervised learning examples

• principal component analysis

• clustering

reinforcement learning

learning what to do, mapping situations to actions, to maximize a numerical reward signal. learner is not told what actions to takakee

reinforcement learning examples

• game AI

• robotics

• LLMs

supervised learning

training set is used to fit the model and test set is used to evaluate model performance

supervised learning assumptions

independent observations and consistent data generating process across both training and test sets

goal of supervised learning

learn a function from a labeled training set (where X and Y are known) that accurately predicts outcomes on unseen data (where only X is known)

spam detection

challenge lies in identifying features that consistently distinguish spam across both seen and unseen emails, while avoiding overfitting to peculiarities in the training data

traditional approach

human experts manually creates rules based on domain knowledge

modern approach

system automatically discovers patterns from labeled examples

ML vs econometrics

ML more focused on correlation and observation, econometrics focused on causal effect and intervention