CS 4803 / 7643 Deep Learning

Spring 2021, WF 2:00 - 3:15 pm, Remote

Course Information

This is an exciting time to be studying (Deep) Machine Learning, or Representation Learning, or for lack of a better term, simply Deep Learning!

Deep Learning is rapidly emerging as one of the most successful and widely applicable set of techniques across a range of domains (vision, language, speech, reasoning, robotics, AI in general), leading to some pretty significant commercial success and exciting new directions that may previously have seemed out of reach.

This course will introduce students to the basics of Neural Networks (NNs) and expose them to some cutting-edge research. It is structured in modules (background, Convolutional NNs, Recurrent NNs, Deep Reinforcement Learning, Deep Structured Prediction). Modules will be presented via instructor lectures and reinforced with homeworks that teach theoretical and practical aspects. The course will also include a project which will allow students to explore an area of Deep Learning that interests them in more depth.

Instructor



Zsolt Kira

Teaching Assistants



Manas Sahni (Head TA)

Andrew Messing

Hrishikesh Kale

Man Xie

Michael Piseno

Rahul Duggal

Sarah Wiegreffe

Yihao Chen


Class meets
Wednesday, Friday 2:00 - 3:15 pm; Remote
Class link
https://primetime.bluejeans.com/a2m/live-event/wfzwvgvv
Piazza
https://piazza.com/class/kjsselshfiz18c (Code: DL2021)
Canvas
CS4803: gatech.instructure.com/courses/172518
CS7643: gatech.instructure.com/courses/172536
Gradescope
CS4803: gradescope.com/courses/228228 (Code: 2RPWYG)
CS7643: gradescope.com/courses/229744 (Code: 4PG38J)

Schedule

Date Topic Optional Reading
W1: Jan 15 Intro lecture + class logistics.
PS0 out: PDF ZIP
Slides (pdf)
W2: Jan 20 Supervised Learning, Linear Classification, Loss functions
PS0 due
Slides (pdf)
W2: Jan 22 Gradient Descent, Linear Algebra View.
PS/HW1 Out
Slides (pdf)
W3: Jan 27 Gradient Descent, Neural Networks.
Slides (pdf)
W3: Jan 29 Neural Networks and Backpropagation.
Slides (pdf)
W4: Feb 3 Neural Networks, Backpropagation, and Automatic Differentiation.
Slides (pdf) [Updated]
Gradients notes and MLP/ReLU Jacobian notes.
W4: Feb 5 Jacobians, AutoDiff, Activation Functions.
Slides (pdf)
W5: Feb 10 Optimization, Regularization, Data Augmentation.
OMSCS Video: Data Wrangling (see dropbox link piazza @8 M1L4 directory).
PS/HW1 due Feb 10th 11:59pm, PS/HW2 out
Slides (pdf)
W5: Feb 12 Optimization, Imbalance, Convolution
Slides (pdf)
W6: Feb 17 Convolution gradients.
Slides (pdf)
CNNs notes, CNNs Backprop notes.
W6: Feb 19 Convolutional Neural Networks (CNNs)
Slides (pdf) (NOTE: Watch M2L6 in dropbox, see piazza)
W7: Feb 24 Visualization
Slides (pdf)
W7: Feb 26 Visualization cont. Advanced CNN architectures.
Slides (pdf)
PS/HW2 due Feb. 28th, PS/HW3 out
W8: Mar 3 Advanced CNN architectures, Bias and Fairness Video (M2L10 on dropbox)
Slides (pdf)
W8: Mar 5 Bias and Fairness, Introduction to Structured Neural Representations, Language Models
Slides (pdf)
W9: Mar 10 Recurrent Neural Networks 2, Long Short-Term Memory
Slides (pdf)
W9: Mar 12 Neural Attention Models (Transformers). Guest Lecture by Arjun Majumdar
Slides (pdf).
PS/HW3 due Mar. 18th, PS/HW4 out
W10: Mar 17 Project Planning session
W10: Mar 19 Masked Language Model details, Embeddings.
Slides (pdf).

Project Proposal due (March 22nd)
W11: Mar 24 No Class (Spring Break Day)
W11: Mar 26 RL background.
Slides (pdf).
Annotated Slides (pdf).
W12: Mar 31 RL: Dynamic Programming (Policy Iteration), Q-Learning, DQN.
Slides (pdf).
W12: Apr 2 RL: Policy Gradients, REINFORCE, Actor-Critic.
Slides (pdf).
PS/HW4 due Apr 8th
W13: Apr 7 Guest Lecture by Michael Auli (FB): Self-supervised learning for speech processing
Slides (pdf).
W13: Apr 9 Guest Lecture by Ishan Misra (FB): Self-Supervised Learning
Slides (pdf).
W14: Apr 14 Guest Lecture by Mahaveer Jain (FB): RNN-T Based Automatic Speech Recognition Systems
Slides (pdf), Source Slides (pptx).
W14: Apr 16 Generative Adversarial Networks (GANs).
Slides (pdf),
W15: Apr 21 VAEs 1.
Slides (pdf),
W15: Apr 23 Few-shot learning, architecture search, wrap-up.
Slides (pdf),
Project due on May 3rd

Grading

  • 80% Homework (4 homeworks)
  • 20% Final Project
  • 3% (potential bonus) Class Participation

Late policy for deliverables

  • No penalties for medical reasons or emergencies. Please see GT Catalog for rules about contacting the office of the Dean of Students. DO NOT send us any private health or other information, and make sure to contact them as soon as you are able after the emergency.
  • All assignments are due as mentioned on Canvas/Piazza. Every homework and project deliverable will have a 48-hour grace period (EXCEPT FOR PS0) during which no penalty will apply. This is intended for you as time to verify that your submission has been submitted (we recommend you re-download it and look it over to make sure all questions/deliverables have been answered). Canvas will show your submission as late, but you do not have to ask for this grace period. Deliverables after the grace period will receive a grade of 0.
  • To re-emphasize: Any assignments submitted after grace period will receive a 0 (no exceptions).

Prerequisites

CS 4803/7643 should not be your first exposure to machine learning. Ideally, you need:

  • Intro-level Machine Learning
    • CS 3600 for the undergraduate section and CS 7641/ISYE 6740/CSE 6740 or equivalent for the graduate section.
  • Algorithms
    • Dynamic programming, basic data structures, complexity (NP-hardness)
  • Calculus and Linear Algebra
    • positive semi-definiteness, multivariate derivates (be prepared for lots and lots of gradients!)
  • Programming
    • This is a demanding class in terms of programming skills.
    • HWs will involve Python and a mix of libraries (PyTorch, NumPy).
    • Your language of choice for project.
  • Ability to deal with abstract mathematical concepts

Project Details (20% of course grade)

The class project is meant for students to (1) gain experience implementing deep models and (2) try Deep Learning on problems that interest them. The amount of effort should be at the level of one homework assignment per group member (2-5 people per group, 3 recommended).

FAQs

  • What is the delivery format?

    Due to Covid-19, this Spring 21 instance is “Remote Sync”, meaning that lectures are delivered over VC at the scheduled class time. All other deliverables are accepted online. The class has no in-person interaction.

    We will be following these Covid-19 guidelines and procedures, with the exception of homework submission, as all homeworks for this course are submitted through Gradescope.

  • The class is full. Can I still get in?

    Sorry. The course admins in CoC control this process. Please talk to them.

  • Unregistered Students who intend to register:

    Whether you are registered or plan to register, if you want to take the course I expect you to attend the classes and submit any deliverables that are assigned in the first or second weeks of class.

    If you are not registered for this course, you will not have access to Gradescope for submission of PS0. Please email the instructor (zkira at gatech) to be added to Gradescope and be able to submit anything due on the first day of class. Make sure to mention which course (CS 4803DL or 7643) you plan to enroll in.

  • Registered students who are not able to access Gradescope:

    This will happen if you were registered to the course very recently. Gradescope rosters are synced periodically and it may take some time for you to receive a Gradescope sign-up notification. If you still face problems with accessing Gradescope, please email the course email (below).

  • I am graduating this Fall and I need this class to complete my degree requirements. What should I do?

    Talk to the advisor or graduate coordinator for your academic program. They are keeping track of your degree requirements and will work with you if you need a specific course.

  • Can I audit this class or take it pass/fail?

    No. Due to the large demand for this class, we will not be allowing audits or pass/fail. Letter grades only. This is to make sure students who want to take the class for credit can.

  • I have a question. What is the best way to reach the course staff?

    Registered students – your first point of contact is Piazza (so that other students may benefit from your questions and our answers). You can also use private messages as well for things that might include answers or other information you do not want others to have (do not send private information such as health information though). If you have a personal matter, email the instructor (zkira at gatech)</p>

Related Classes / Online Resources

Book

Overviews

Note to people outside Georgia Tech

Feel free to use the slides and materials available online here. If you use our slides, an appropriate attribution is requested. Please email the instructor with any corrections or improvements.