The Machine Learning Theory Reading Group (MLTRG) meets regularly (usually weekly) to talk about recent research. In terms 1 and 2, we focus on a particular topic. In the summer, we discuss a variety of papers with less cohesiveness.
To receive announcements about the reading group, please join our mailing list by sending an email to majordomo@cs.ubc.ca with "subscribe mltrg-l YOUR-EMAIL-ADDRESS" in the message body.
| Summer 2017 (The Algorithmic Foundations of Differential Privacy by Dwork and Roth) Every Tuesday in room ICICS 304 from 2-4pm | ||
| Date | Presenter | Topic |
| May 9 | Chris | Chapter 3: Basic Techniques and Composition Theorems (Laplace and Exponential Mechanisms) |
| May 16 | Chris | Chapter 3: Basic Techniques and Composition Theorems (Composition and sparse vector technique) |
| May 23 | Sikander | Chapter 4: Releasing linear queries with correlated eror |
| May 30 | Devon | Chapter 5: Generalizations |
| June 6 | Aaron | Chapter 6: Boosting for queries |
| June 13 | Raunak | Chapter 7: When worst-case sensitivity is atypical |
| Winter 2017 Every Wednesday in room ICICS 304 from 1-3pm | ||
| Date | Presenter | Topic |
| Jan. 11 | Chris | Rademacher complexity |
| Jan. 18 | Devon and Raunak | VC Dimension |
| Jan. 25 | Bader | On the mathematical foundations of learning |
| Feb. 1 | Bader | On the mathematical foundations of learning, continued |
| Feb. 8 | Bader | On the mathematical foundations of learning, continued |
| Feb. 22 | Issam | Support Vector Machines |
| Mar. 15 | Sharan | Train faster, generalize better: Stability of stochastic gradient descent |
| Mar. 22 | Sikander | Nonuniform Learning |
| Apr. 19 | Issam | Neural Network Theory |
| Fall 2016 (Deep learning) Every Monday in room ICICS 304 from 3:30-5:30pm | ||
| Date | Presenter | Topic |
| Sept. 12 | Chris | Learning neural networks |
| Sept. 19 | Jason | The loss surfaces of multilayer networks (spin glass) |
| Sept. 26 | Abbas | Benefits of depth in neural networks |
| Oct. 3 | Nick | Sketching and neural networks |
| Oct. 17 | Reza | Convexified convolutional neural networks |
| Oct. 24 | Ricky | What regularized auto-encoders learn from the data generating distribution |
| Oct. 31 | Chris | Beating the perils of non-convexity: guaranteed training of neural networks using tensor methods |
| Nov. 7 | Saif | Universal approximation theorem (Cybenko, Hornik et al.) |
| Nov. 14 | Issam | AdaNet |
| Nov. 21 | Julie | Finding approximate local minima for nonconvex optimization in linear time |
| Nov. 28 | Free discussion | Talk about ideas, interesting open questions, directions for further research, etc. |
| Summer 2016 Every Tuesday in room ICICS 304 from 1-3pm | ||
| Date | Presenter | Topic |
| May 10 | Reza | Hazan's Book, Chapter 1: Introduction and Chapter 2: Convergence analysis in convex optimization |
| May 17 | Chris | Hazan's Book, Chapter 3: First order algorithms for online convex optimization |
| May 24 | Sharan | Hazan's Book, Chapter 4: Second order methods |
| June 7 | Abbas | Hazan's Book, Chapter 5: Regularization |
| June 14 | Abbas | Hazan's Book, Chapter 5: Regularization continued |
| June 21 | Sharan | Hazan's Book, Chapter 6: Bandit convex optimization |
| June 28 | Sharan | Hazan's Book, Chapter 6: Bandit convex optimization, continued |
| July 5 | Chris | Streaming SVRG |
| July 12 | Abbas | Local Superlinearly-convergent method for finite sums, ICML'16 |
| July 19 | Mark | Katyusha: The First Direct Acceleration of Stochastic Gradient Methods |
| Winter 2016 Every Monday in room ICICS 304 from 3-5pm | ||
| Date | Presenter | Topic |
| Feb 13 | Abbas and Chris | Moitra's Book, Chapter 2: Non-negative Matrix Factorization |
| Feb 29 | Sharan and Reza | Moitra's Book, Chapter 3: Tensor Methods |
| Feb 29 | Abbas and Yaniv | Moitra's Book, Chapter 4: Sparse Recovery |
| March 7 | Chris | Moitra's Book, Chapter 5: Dictionary Learning |
| March 14 | Sharan and Reza | Moitra's Book, Chapter 6: Mixtures of Gaussians |
| March 21 | Abbas and Chris | Moitra's Book, Chapter 7: Matrix completion |
| Apr 4 | Abbas | Iterative methods for matrix completion, based on a STOC'13 and a COLT'15 paper |
| Apr 11 | Nick | Circuits of the Mind and Cortical Learning, COLT'15 |
| Apr 18 | Sharan | Some theory of Deep Learning, based on an ICLR'16 paper |
| Apr 25 | Mark | Overview of stochastic gradient methods |