Introduction to Supervised Learning
STAT 195
Subject & Catalog Number
Course Information
Description
This course will dive into the statistical properties of various machine learning methods that arise in supervised problems. The goal of the course is to introduce and prepare students for theoretical and methodological research in statistical machine learning. Topics we will discuss include nearest neighbors, no free lunch theorems, curse of dimensionality, structured learning, subset selection, shrinkage methods, principal components regression, double descent, implicit regularization and interpolation, optimism, effective number of parameters, cross validation, ensembling, transfer learning, multimodal learning, predictive inference, algorithmic fairness, causal inference using machine learning. In addition to problem sets but instead of exams, students will work in groups to synthesize and give short presentations on recent machine learning papers. Building upon these presentations, students will conduct individual course projects, based on which they will submit a final report and give either a presentation/poster presentation at the end of the semester.
Course Notes
We will release lectures notes. For additional reading, the main reference will be Elements of Statistical Learning by Hastie, Tibshirani, and Friedman, which is available for free at https://web.stanford.edu/~hastie/ElemStatLearn/printings/ESLII_print12.pdf.
Recommended Prep
Required: CS 181 or CS/Stat 109A, and Stat 110. Recommended: Stat 111.
Available for Harvard Cross Registration
NOTE: This course requires additional sections; you will be prompted to choose secondary components during the Add to Cart process
