AI for Statistical Learning

STAT 6393V-001: Topics Course in Statistics

Dates: M/T/W/T/F 1245-1415 in Summer 1

Prerequisites: None; any undergrad level stats course will be a bonus

Textbook: (Recommended) Bishop & Bishop. Deep Learning: Foundations and Concepts. link

Software tools: R, Google Colab, Git

Course Description

This is an intermediate course on using neural networks for statistical learning using the R programming language. Neural networks/deep learning are the most important class of models in the field of artificial intelligence; the modern technology stack relies heavily on it, and it is ubiquitous in both academia and industry, from weather prediction models to spam filters to personalized medicine. This course explores the synergy between statistics and deep learning and how a) neural networks can augment classical statistical techniques for modeling complex phenomena, and b) statistical inference can make AI models more robust and explainable.

Students will be provided a crash course in distribution theory, statistical inference, and R. Details regarding loss functions and optimization will be covered which are widely applicable to any modern machine learning approach. Students will also be introduced to advanced topics like Bayesian neural networks, simulation-based inference, and uncertainty quantification.

An AI4Stats short course conducted in Fall 2025 at the University of Maryland, Baltimore County, covered several of the ideas that underpin this course, but tailored towards an audience that was already familiar with statistics, or deep learning, or both.

Course Schedule

Week 1: Probability and distribution theory; frequentist and Bayesian methods

Week 2: Loss functions and numerical optimization; building deep learning models in R

Week 3: Custom loss functions; regression and classification; parameter and density estimation

Week 4: Bayesian neural networks; conformal prediction; (deep) Gaussian processes

Week 5: Simulation-based inference; generative modeling