CS 747
Deep Learning

Time/Location: Tuesday 4:30-7:10pm, Planetary Hall 120  
Instructor: Jana Kosecka
Office hours:  4444 ENGR 2-3pm Monday
TA:  Xue Yu, xyu21@gmu.edu, Office hours 1:30-2:30pm Tuesday, ENGR 4456
Contact: Office 4444 Engineering Building, e-mail: kosecka@gmu.edu, 3-1876
Resources, schedule, handouts: piazza.com/gmu/fall2022/cs747

This course will cover an introduction to neural networks and deep learning. We will cover multi-layer neural networks, convolutional neural networks, recurrent neural networks, transformers, generative neural networks and deep reinforcement learning. We will discussed representative models and techniques for image classification, image and text generation, perception and action.The class will consist of programming assignments in Python (PyTorch), paper review/presentations and final project.

The course will comprise of lectures by the instructor, homeworks, paper review/presentations and final project.

Prerequisites:

CS 688, strong programming experience and willingness to participate in discussions
Students taking the class should be comfortable with linear algebra, calculus and probability

Recommended Textbook:

Deep Learning by Goodfellow et. link here
Deep Learning with Python, 2nd edition by F. Chollet
Dive into Deep Learning" (online available: https://d2l.ai/)

Grading:

Assignments: 50%
Paper summaries, activities, presentations: 15%
Class participation: 5%
Final project: 30%

Grading scale:

A   >93
A-   90-93
B+   87-90
B   83-87
B-   80-83
C+   77-80
C   72-77
C-   67-72
D   60-67
F   < 60

Late policy: Each student will have a 5 day late submission budget, which could be used towards late submission on the homeworks.

Outlines of topics:

  • Machine learning refresher, Python/numpy, linear classifiers
  • Neural Networks, Backpropagation, Computational Graphs
  • Optimization, common loss functions, training neural networks
  • Convolutional neural networks, object detection, dense prediction
  • Autoencoders, Variational Autoencoders Generative Adversarial Networks
  • Recurrent Neural Networks, Attention, Transformers
  • Metric and unsupervised learning, adversarial examples
  • Deep Reinforcement Learning
  • Applications: Computer Vision, Robotics, Natural Language Processing and others

    CS department Honor Code can be here.

    Disability Statement If you have a documented learning disability or other condition that may affect academic performance you should: 1) make sure this documentation is on file with the Office of Disability Services (SUB I, Rm. 222; 993-2474; www.gmu.edu/student/drc) to determine the accommodations you need; and 2) talk with me to discuss your accommodation needs