18-739: Security and Fairness of Deep Learning: Spring 2020

Classes begin January 14. See CMU Academic Calendar.

All homework is due 10 minutes before lecture start.

Date	Topic	Reading
Part I: Background
Tuesday January 14	Slides Course Overview CMU Computing Policy CMU Policy on Academic Integrity Nature Review on Deep Learning Classification Intro
Thursday January 16	Slides Classification, Logistic Regression Deep Learning Book, CH2 Linear Classification Basics
Tuesday January 21	Slides Gradient Descent, Stochastic Gradient Descent Stochastic Gradient Descent Deep Learning Book, CH3
Thursday January 23	Homework 1 out (pdf , zip) Slides Tutorial GCP Setup Deep Learning Software Keras Tensorflow Numpy
Tuesday January 28	Slides Back Propagation Back Propagation
Thursday January 30	Slides Deep Feedforward Networks Deep Learning Book, CH6 Neural Networks Setting up the Data and Model
Tuesday February 4	Slides Deep Feedforward Training Deep Learning Book, CH6 Learning and Evaluation
Thursday February 6	Homework 1 due Slides Background: Convolutional Neural Networks I CNN Architecture Deep Learning Book, CH9 Image Kernels Demo
Tuesday February 11	Slides Background: Second Order Optimization CNN Visualization Deep Learning Book, CH9
Part II: Explanations
Thursday February 13	Homework 2 out (pdf , zip) Slides Paper Discussion: Representer Point Selection for DNN Representer point selection for DNN
Tuesday February 18	Slides Paper Discussion: Black-box Predictions Understanding Black-box Predictions via Influence Functions
Thursday February 20	Slides Paper Discussion: Axiomatic Attribution for Deep Networks Axiomatic Attribution for Deep Networks
Tuesday February 25	Slides Paper Discussion: Influence-Directed Explanations for CNNsGuest speaker: Klas Leino Influence-Directed Explanations for CNNs
Thursday February 27	Slides Explanations Review (optional) "Why Should I Trust You?": Explaining the Predictions of Any Classifier (optional) An Evaluation of the Human-Interpretability of Explanation
Tuesday March 3	Homework 2 due Slides Paper Discussion: Shapley Values and Cooperative Game TheoryGuest speaker: Ankur Taly The Explanation Game: Explaining Machine Learning Models with Cooperative Game Theory (optional) The many Shapley values for model explanation
Thursday March 5	Homework 3 Part 1 out (zip+pdf) Slides Paper Discussion: Concept-based ExplanationsGuest speaker: Amirata Ghorbani Towards Automatic Concept-based Explanations (optional) Quantitative Testing with Concept Activation Vectors (TCAV)
Tuesday March 10	No Class: Spring Break
Thursday March 12	No Class: Spring Break
Part III: Adversarial Learning
Tuesday March 17	Homework 3 Part 2 out (see canvas) Teleconferencing Debugging Session and Office Hours
Thursday March 19	Slides Paper Discussion: Adversarial LearningGuest speaker: Nicolas Papernot The Limitations of DL in Adversarial Settings
Tuesday March 24	Paper Discussion: Real-world Adversarial AttacksGuest speaker: Mahmood Sharif DReal and Stealthy Attacks on State-of-the-Art Face Recognition (optional) Face Recognition and Privacy
Thursday March 26	Slides Paper Discussion: Robustness and Adversarial LearningGuest speaker: Nicholas Carlini Towards Evaluating the Robustness of Neural Networks
Tuesday March 31	Homework 3 due Slides Paper Discussion: Generative Adversarial Nets Generative Adversarial Nets
Part IV: Privacy and Fairness
Thursday April 2	Homework 3 makeup out Paper Discussion: Whitebox Inference AttacksGuest speaker: Milad Nasr Comprehensive Privacy Analysis of Deep Learning: Stand-alone and Federated Learning under Passive and Active White-box Inference Attacks
Tuesday April 7	Slides Paper Discussion: Resistance to Adversarial AttacksGuest speaker: Dimitris Tsipras Towards Deep Learning Models Resistant to Adversarial Attacks
Thursday April 9	Homework 3 makeup due Slides Paper Discussion: Fairness in Deep Learning Mitigating Unwanted Biases with Adversarial Learning
Tuesday April 14	Slides Background: Recurrent Neural Networks I Deep Learning Book, CH10 word2vec
Thursday April 16	Homework 4 Part 1 out Slides Background: Recurrent Neural Networks II Deep Learning Book, CH10 Sections 10.1 and 10.2 N-gram Language Models The Unreasonable Effectiveness of Recurrent Neural Networks
Tuesday April 21	Paper Discussion: Bias In Word EmbeddingsGuest speaker: Kai-Wei Chang Man is to Computer Programmer as Woman is to Homemaker?
Thursday April 23	Homework 4 Part 1 due (approximate date) Homework 4 Part 2 out Paper Discussion: Bias in NLP tasksGuest speaker: Caleb Lu Gender Bias in Neural Natural Language Processing
Tuesday April 28	Paper Discussion: Real-world Adversarial AttacksGuest speaker: Mahmood Sharif DReal and Stealthy Attacks on State-of-the-Art Face Recognition (optional) Face Recognition and Privacy
Thursday April 30	(approximate date) Homework 4 Part 2 due (approximate date) Homework 5 Part 1 out TBD

* All schedules are subject to change over the course.