Topics in Statistics (359-0-20)

Topic

Deep Learning

Instructors

Arvind Krishna

Meeting Info

Harris Hall 107: Tues, Thurs 5:00PM - 6:20PM

Overview of class

Topic: Deep Learning
This course introduces state-of-the-art deep learning architectures, such as CNN, RNN, LSTM, transformers, autoencoders, adversarial networks, and their applications across domains such as large language models, fraud detection, speech recognition, and medical imaging, and autonomous systems. Students will learn the theoretical foundations of these models, understand their distinct purposes, and gain practical experience applying them to real-world problems, while also engaging with recent advances in deep learning research. 

Registration Requirements

STAT 362-0 (or an equivalent machine learning course), MATH 240-0 (liner algebra), MATH 230-2 or STAT 228-0 or MATH 235-0 (calculus)

Learning Objectives

On completion of the course, students will be able to:
Have a solid conceptual understanding and theoretical foundation of the current state-of-the-art deep learning methods
Interpret latent representations and embeddings
Distinguish between different architectures and when to use them
Train, evaluate, test, diagnose, and fix deep learning models with the python programming language

Teaching Method

The instructor will lead the class through lectures, with a strong emphasis on conceptual and theoretical understanding. Students are encouraged to actively participate by asking questions, and the instructor will further promote engagement through in-class conceptual prompts. Coding support will primarily be offered during office hours. However, if common coding challenges or important implementation nuances arise—particularly related to specific methods or assignments—they will be addressed during class sessions.

Evaluation Method

Students will be assessed on the learning objectives with:
1. Weekly assignments: Students will have bi-weekly assignments to demonstrate their ability to train, evaluate, test, diagnose, and fix deep learning models, and analyze results.
2. Mid-term exam: Students will have a mid-term exam, where they will be assessed on their conceptual and theoretical understanding.
3. Final exam: Students will complete a final exam designed to assess their conceptual and theoretical understanding, as well as their ability to select, train, evaluate, and analyze appropriate deep learning models for an open-ended problem.
4. Literature survey: Students will be assessed on their ability to read and critically evaluate recent deep learning papers.

Class Materials (Required)

Deep Learning - By Ian Goodfellow, Yoshua Bengio, and Aaron Courville
Free online: https://www.deeplearningbook.org/

Class Materials (Suggested)

Links to research papers / articles will be provided during the course

Enrollment Requirements

Enrollment Requirements: Preregistration for this class is reserved for Statistics majors and minors as well as Data Science majors.
Add Consent: Instructor Consent Required