Deep Learning
Logistics
When: Tuesday and Thursday, 11:00 am-12:15 pm ET
Where: SN 011 (Sitterson Hall)
TA Office hours: TBA
Instructor Office hours: 12:15 -1:15 Thu (email for appt)
Sign up on the course Piazza here.
People
Course Information
Deep Learning (the field of training neural networks) has transformed the landscape of AI in the last decade, and driven successful advances in numerous domains including language technologies, computer vision, and many others. This course provides an in-depth introduction to deep learning. It covers fundamental concepts and state-of-the-art methodologies in building and optimizing neural networks, emphasizing hands-on experience with different model architectures and training mechanisms. The course begins with a review of linear models, and then describes feedforward neural nets, CNNs, RNNs and Transformer-based Architectures. The course ends with an overview of some other generative AI methods, and a discussion of ethical considerations in deep learning systems. A tentative list of topics follows.
Linear & Logistic Regression
Feedforward networks & backpropagation
Regularization & Over/under-fitting
Convolutional Neural Networks
Representation Learning and Word Embeddings
Sequence Modeling, Seq-2-Seq and RNNs
Attention & Transformer-based architectures
Generative Models
Ethics in Deep Learning
Prerequisites
Machine learning (one of COMP562, COMP755, STOR565, or equivalent)
Linear algebra (one of MATH210, MATH347, MATH577, or equivalent)
Multivariate calculus (one of MATH233, MATH522, or equivalent)
The course will presume a functional understanding of Machine Learning and Linear algebra. The assignments and project will have substantial programming components.
Class Schedule (tentative)
Grading
The class grade will have these components:
Weekly Quizzes (20%): Quizzes will have MCQ-type questions, and will test the understanding of each week's topics.
Assignments (40%) : Assignments will have a mix of coding and math.
Midterm exam (40%): This will be an in-class exam, and cover all topics covered before the midterm. The tentative date for the exam is Oct 24.