CS 7150: Deep Learning - Summer-Full 2020

Time & Location:

3:20 - 5:00pm Eastern Time, Mondays, Location: Online. Note: Access details have been posted on the Canvas page for the course.

Staff

Instructor: Paul Hand
Email: p.hand@northeastern.edu Office Hours: Online by appointment

TA: Danny Goldfarb
Email: goldfarb.d@husky.neu.edu Office Hours: Online by appointment

Course Description

Note: This differs from the official course description. Please read carefully.

Introduction to deep learning, including multi-layer perceptrons, convolutional neural networks, adversarial examples, autoencoders, U-nets, Residual Networks, variational autoencoders, generative adversarial networks, gradient-based optimization methods, back propagation, automatic differentiation, and applications. Assumes students already have a basic knowledge of machine learning, optimization, and statistics.

Overview

The learning objectives of this course are that students should:

be able to train neural networks at a variety of tasks
be able to select appropriate neural network architectures for a variety of tasks
have read and be able to discuss the contents and contributions of important papers in the field
have a basic theoretical understanding of the tools of deep learning

Course Structure:

This course will be conducted entirely online. Mondays will consist of a synchronous all-class video conference where we will discuss the assigned papers in detail. Additionally, you will be expected to view asynchronous lectures that will be posted to the internet. You may watch these videos anytime after they are posted. Office hours are by appointment and will be held online.

Student Work:

Students will be expected to complete the following work:

Each week, students will be given a set of questions to answer about that weeks reading assignment. Students are expected to submit the answers to these questions to Gradescope by noon on Mondays in order to ensure they are prepared to engage in the class discussion.
Students are expected to participate in the whole-class discussions. Each student is expected to contribute to the whole-class discussion at least twice over every 3 week period.
There will be three homework assignments throughout the semester. They will involve both pencil-and-paper work and computation involving training neural networks. The results will be submitted to gradescope in the form of a pdf.
Students will complete a project. They will reproduce an empirical observation in a paper of their choice by independently recreating and executing code to train and test that network.

Student work will be graded as follows: 0 - Missing, Absent, Inadequate, Disengaged; 1 - Engaged, Thoughtful, Accurate. This includes the paper question responses, participation in the whole-class discussions, each homework problem, and the project.

Course grades:

Course grades will be based on: 30% Paper questions, 30% Participation, 30% HWs, 10% Project.
Letter grades will be assigned on the following scale: 93%+ A, 90-92% A-, 87-89 B+, 83-87% B, 80-82% B-, 77-79 C+, 73-77% C, 70-72% C-,60-70% D, 0-59% F.

Online Etiquette:

During class, students should be fully engaged with the class, to the best of their ability. Students are expected to leave their video connection on for the majority of class. If this is not possible, please consult with the instructor and an accomodation will be made on a case by case basis. Participants should mute themselves when they are not speaking. Participants should willingly volunteer to answer questions and should be prepared to be called upon to answer questions if needed.

Prerequisites

Students are expected to have experience with a class in Machine Learning. The class will assume students are comfortable with linear algebra and probability. Some experience with neural networks, python, PyTorch/TensorFlow will be helpful but can be acquired while taking this class.

Week	Date	Class Discussion Will Be On:	Lecture to Watch Before Class
1	M 5/4	A DARPA Perspective on Artificial Intelligence Preparation Questions for Class
2	M 5/11 HW 1 Released (tex)	Deep Learning Visualizing and Understanding Convolutional Networks Preparation Questions for Class (tex)	Architectural Elements of Neural Networks. (Notes).
3	M 5/18	Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift How Does Batch Normalization Help Optimization? Video of Ian Goodfellow explaining batch normalization Preparation Questions for Class (tex)	Machine Learning Review (Notes)
-	M 5/25 HW 1 DUE 5/27	No Class - Memorial Day
4	M 6/1 HW 2 Released (tex)	Deep Residual Learning for Image Recognition (ResNets) ImageNet Classification with Deep Convolutional Neural Networks (AlexNet) Preparation Questions for Class (tex)	Neural Network Architectures for Images (Notes)
5	M 6/8	No Class - University-wide Day of Reflection, Engagement, and Action
6	M 6/15	Explaining and Harnessing Adversarial Examples Robust Physical-World Attacks on Deep Learning Models Preparation Questions for Class (tex)	Adversarial Examples for Deep Neural Networks (Notes)
7	M 6/22 HW 2 DUE 6/24 HW 3 Released (tex)	Overcoming catastrophic forgetting in neural networks Preparation Questions for Class (tex) Notes from class	Continual Learning and Catastrophic Forgetting (Notes)
8	M 6/29	Learning From Noisy Singly-labeled Data Preparation Questions for Class (tex) (Notes from 6/29 class) (Notes from 7/6 class)	Gradient Descent and Stochastic Gradient Descent (notes)
9	M 7/6	Project Time Preparation Questions for Class	Backprop and Autograd
10	M 7/13 HW 3 DUE 7/15 Project PLANNING Document Released (tex)	Auto-Encoding Variational Bayes Preparation Questions for Class (tex)	Variational Autoencoders (notes)
11	M 7/20 Project PLANNING Document DUE 7/22	Generative Adversarial Networks Wasserstein GAN Preparation Questions for Class (tex)	Generative Adversarial Networks (notes)
12	M 7/27	Density estimation using Real NVP Preparation Questions for Class (tex)	Invertible Neural Nets (notes)
13	M 8/3 Project DUE 8/3	Project Presentations