Classes start on January 15, 2019
| Date | Topic | Reading | Notes |
|---|---|---|---|
| Tue, Jan 15 | Course Overview |
CMU Computing Policy CMU Policy on Academic Integrity Nature Review on Deep Learning Classification Intro |
Slides |
| PART I: Machine Learning Basics | |||
| Thur, Jan 17 | Intro to Machine Learning I (Linear Algebra Review, Logistic Regression) |
Deep Learning Book, CH2 Linear Classification Basics |
Slides |
| Tue, Jan 22 | Intro to Machine Learning II (Gradient Descent, Stochastic Gradient Descent)) HW1 Out: | Stochastic Gradient Descent Deep Learning Book, CH3 | Slides |
| Thur, Jan 24 | Deep Learning Software | Documentation of Keras Documentation of Tensorflow | Slides |
| PART II: Intro to Deep Learning | |||
| Tue, Jan 29 | Back Propagation |
Back Propagation |
Slides |
| Thur, Jan 31 |
Deep Feedforward Networks |
Deep Learning Book, CH6 Setting up the Data and Model |
|
| Tue, Feb 5 |
DFN Training HW1 Due (before class) |
Deep Learning Book, CH6 Learning and Evaluation |
Slides |
| Thur, Feb 7 | DFN Training HW2 Out |
Deep Learning Book, CH6 Learning and Evaluation |
|
| Tue, Feb 12 |
Convolutional Neural Networks I HW2 Tutorial |
CNN Architecture Deep Learning Book, CH9 | Slides |
| Thur, Feb 14 | Convolutional Neural Networks II | CNN Visualization Deep Learning Book, CH9 | Slides |
| PART III: Explanation for Deep Neural Networks | |||
| Tue, Feb 19 | Paper Discussion: Influence-Directed Explanations for CNNs | Influence-Directed Explanations for CNNs | Slides |
| Thur, Feb 21 | Paper Discussion: Feature-Wise Bias Amplifications HW3 out Friday Feb 22 | Feature-Wise Bias Amplifications | Slides |
| Tue, Feb 26 | Paper Discussion: Feature-Wise Bias Amplifications |
Feature-Wise Bias Amplifications optional: Understanding Black-box Predictions via Influence Functions |
|
| Thur, Feb 28 | Paper Discussion: Chih-Kuan Yeh & Joon Sik Kim | Representer point selection for DNN | Slides |
| Tue, Mar 5 | Paper Discussion: Been Kim |
Quantitative Testing with Concept Activation Vectors (TCAV) Sanity Checks for Saliency Maps An Evaluation of the Human-Interpretability of Explanation |
Slides |
| PART IV: Adversarial Learning | |||
| Thur, Mar 7 | Paper Discussion: Adversarial Settings in Deep Learning | The Limitations of DL in Adversarial Settings | Slides |
| Tue, Mar 12 | No Class: Spring Break | ||
| Thur, Mar 14 | No Class: Spring Break | ||
| Tue, Mar 19 | Paper Discussion: Nicolas Carlini | Towards Evaluating the Robustness of Neural Networks | Slides |
| Thur, Mar 21 |
Paper Discussion: Mahmood Sharif
HW 3 Due before class HW 4 Part I Out |
DReal and Stealthy Attacks on State-of-the-Art Face Recognition Face Recognition and Privacy |
Slide1 Slide2 |
| Tue, Mar 26 | Overview of Privacy Attacks | Reference Papers at the end of the slide | Slide |
| Thur, Mar 28 | Paper Discussion: Whitebox Membership Inference | TBD | Slide |
| PART V: Fairness and Recurrent Neural Networks | |||
| Tue, Apr 2 | Overview of Fairness in DL: Emily Black (Guest Lectuerer) | Mitigating Unwanted Biases with Adversarial Learning | Slide |
| Thur, Apr 4 | Fairness in DL: Emily Black (Guest Lectuerer) | Papers in the reference slide | Slide |
| Tue, Apr 9 | Word Embeddings | word2vec | Slide |
| Thur, Apr 11 | CMU Carnival: No class HW 4 Due (before 10:30 AM EST) |
||
| Tue,Apr 16 | Paper Discussion: Bias In Word Embeddings | Man is to Computer Programmer as Woman is to Homemaker? Human-like Bias in Langauge Models | Slide |
| Thur, Apr 18 | Recurrent Neural Networks I HW 5 Part I Out | Deep Learning Book, CH10 | Slide |
| Tue, Apr 23 | Recurrent Neural Networks II |
Deep Learning Book]: Sequence Modeling: Recurrent and Recursive Neural Nets(Sections 10.1 and 10.2) CH10 N-gram Language Models (textbook chapter) The Unreasonable Effectiveness of Recurrent Neural Networks(blog post overview) |
Slide |
| Thur, Apr 25 | Bias in NLP tasks HW 5 Part II Out | Bias in NLP tasks | Slide |
| Thur, Apr 25 | Recurrent Neural Networks III: LSTM |
Deep Learning Book: Sequence Modeling: Recurrent and Recursive Neural Nets(Sections 10.3, 10.5, 10.7-10.12) Learning long-term dependencies with gradient descent is difficult (one of the original vanishing gradient papers) Exploding and Vanishing Gradients Understanding LSTMs |
Slide1 Slide2 Slide3 |
| Thur, May 2 |
Class Wrapup HW 5 Due May 7th 1:30PM EST |
TBD | |
* Schedule is subject to change.