Course Information
About
This course provides a thorough understanding of the fundamental concepts and recent advances in deep learning. The main objective is to provide students practical and theoretical foundations to use and develop deep neural architectures to solve challenging tasks in an end-to-end manner. The course is taught by Aykut Erdem. The teaching assistant is Emre Can Açıkgöz.
The course will use Deep Learning by Ian Goodfellow, Yoshua Bengio and Aaron Courville as the textbook (draft available online and for free here).
Instruction style: During the semester, students are responsible for studying and keeping up with the course material outside of class time. These may involve reading particular book chapters, papers or blogs and watching some video lectures. After the first three lectures, each week a group of students will present a paper related to the topics of the previous week.
Time and Location
Lectures: Mondays and Wednesday at 08:30-09:40 (SOS 103)
Tutorials: Tuesday at 17:30-18:40 (SNA A44)
Communication
The course webpage will be updated regularly throughout the semester with lecture notes, presentations, assignments and important deadlines. All other course related communications will be carried out through Blackboard.
Prerequisites
COMP541 is open to all graduate students in our COMP department. Prospective senior undergraduate COMP students and non-COMP graduate students, however, should ask the course instructor for approval before the add/drop period. The prerequisites for this course are:
- Programming (you should be a proficient programmer to work out the practicals and to implement your course project.)
- Calculus (differentiation, chain rule) and Linear Algebra (vectors, matrices, eigenvalues/vectors)
- Basic Probability and Statistics (random variables, expectations, multivariate Gaussians, Bayes rule, conditional probabilities)
- Machine Learning (supervised and unsupervised learning, linear regression, overfitting, underfitting, regularization, bias vs variance tradeoff)
- Optimization (cost functions, taking gradients, regularization)
Course Requirements and Grading
Grading for COMP541 will be based on
- Self-Assessment Quiz (2%)
- Assignments (20%) (4 assignments x 5% each)
- Midterm Exam (21%)
- Course Project (presentations and reports) (32%),
- Paper Presentations (10%),
- Paper Reviews (5%),
- Class Participation (10%)
Reference Books
- Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep Learning, MIT Press, 2016 (draft available online)
Schedule
| Date | Topic | Assignments |
| Oct 2 | Introduction to Deep Learning | Self-Assessment Quiz (Theory) |
| Oct 9 | Machine Learning Overview | Self-Assessment Quiz (Programming) |
| Oct 16 | Multi-Layer Perceptrons | Assg 1 out: MLPs and Backpropagation |
| Oct 23 | Training Deep Neural Networks | |
| Oct 30 | Convolutional Neural Networks | Assg 1 due, Assg 2 out: CNNs |
| Nov 6 | Understanding and Visualizing CNNs | Project proposal due |
| Nov 13 | Winter Break | |
| Nov 20 | Recurrent Neural Networks | Assg 2 due, Assg 3 out: RNNs |
| Nov 27 | Attention and Transformers | |
| Nov 30 | Midterm Exam (guide) | |
| Dec 4 | Graph Neural Networks | Assg 3 due, Assg 4 out: Transformers and GNNs |
| Dec 11 | Generative Adversarial Networks | Project progress report due |
| Dec 18 | Autoregressive Models | Assg 4 due |
| Dec 25 | VAEs, Diffusion Models | |
| Jan 1 | VAEs, Diffusion Models (cont'd.) | |
| Jan 8 | Self-supervised Learning | |
| Jan 22 | Final Project Presentations | Final project due |