Syllabus for CS6787

Advanced Machine Learning Systems — Fall 2017

TermFall 2017InstructorChristopher De Sa
Course[email hidden]
ScheduleMW 7:30pm – 8:45pmOffice hoursW 2:00pm – 3:00pm or by appointment
RoomBill and Melinda Gates Hall G01OfficeBill and Melinda Gates Hall 450

Description: So you've taken a machine learning class. You know the models people use to solve their problems. You know the algorithms they use for learning. You know how to evaluate the quality of their solutions.

But when we look at a large-scale machine learning application that is deployed in practice, it's not always exactly what you learned in class. Sure, the basic models, the basic algorithms are all there. But they're modified a bit, in a bunch of different ways, to run faster and more efficiently. And these modifications are really important—they often are what make the system tractable to run on the data it needs to process.

CS6787 is a graduate-level introduction to these system-focused aspects of machine learning, covering guiding principles and commonly used techniques for scaling up to large data sets. Informally, we will cover the techniques that lie between a standard machine learning course and an efficient systems implementation. Topics will include stochastic gradient descent, acceleration, variance reduction, methods for choosing metaparameters, parallelization within a chip and across a cluster, popular ML frameworks, and innovations in hardware architectures. An open-ended project in which students apply these techniques is a major part of the course.

Prerequisites: Knowledge of machine learning at the level of CS4780. Optionally, knowledge of computer systems and hardware on the level of CS 3410 would be useful, but this is not a prerequisite.

Format: For half of the classes, typically on Mondays, there will be a traditionally formatted lecture. For the other half of the classes, typically on Wednesdays, we will read and discuss a seminal paper relevant to the course topic. These classes will involve a presentation by a group of students of the paper contents (each student will sign up in a group to present one paper) followed by breakout discussions about the material.

Grading: Students will be evaluated on the following basis.

20%Paper presentation
10%In-class quizzes — there will be a quiz before each paper presentation on that paper's content
10%Discussion participation
30%Paper reviews — students must submit a review of every paper we discuss
30%Final project

Paper review parameters: Paper reviews should be about one page (single-spaced) in length. The review guidelines should mirror what an actual conference review would look like (although you needn't assign scores or anything like that). In particular you should at least: (1) summarize the paper, (2) discuss the paper's strengths and weaknesses, and (3) discuss the paper's impact. For reference, you can read the NIPS reviewer guidelines, starting with the Overview section on page 6. Of course, your review will not be precisely like a real review, in large part because we already know the impact of these papers.

Final project parameters and course calendar may be subject to change.

Final project parameters: [Project Overview Slides] The final project can be done in groups of up to three (although more work will be expected from groups with more people). The subject of the project is open-ended, but it must include:

Project proposals are due on Monday, November 13. The project proposal should satisfy the following constraints: The project will culminate in a project report of at least four-pages, not including references. The project report should be formatted similarly to a workshop paper, and should use the ICML 2017 style or a similar style. An abstract for the report is due on Monday, November 27, and we will discuss the abstracts in class on that day. The final project report is due on Wednesday, December 6.

Course Calendar

Wednesday, August 23 No in-person lecture. I am traveling this week. Do not go to the lecture room. No one will be there.
Monday, August 28 Lecture 1. [Slides] [Demo Notebook] Topics:
  • Overview
  • Course outline and syllabus
  • Gradient descent
  • Stochastic gradient descent: the workhorse of machine learning
  • Theory of SGD for convex objectives
Wednesday, August 30 Due: Sign-up for paper presentations.

Lecture 2. [Slides] [Demo Notebook] Topics:
  • The effect of choosing the step size/learning rate
  • Mini-batching and batch size
  • Overfitting
  • Generalization error
  • Regularization
  • Early stopping
Monday, September 4 Labor day. No lecture.
Wednesday, September 6 Paper Discussion 1. Rich Caruana, Steve Lawrence, and C Lee Giles. Overfitting in neural nets: Backpropagation, conjugate gradient, and early stopping. In Advances in neural information processing systems, pages 402–408, 2001
Monday, September 11 Due: Review of Paper 1.

Lecture 3. [Slides] [Demo Notebook] Topics:
  • The condition number
  • Momentum and acceleration
  • Momentum for quadratic optimization
  • Momentum for principle component analysis (PCA)
Wednesday, September 13 Paper Discussion 2. Ilya Sutskever, James Martens, George Dahl, and Geoffrey Hinton. On the importance of initialization and momentum in deep learning. In International conference on machine learning, pages 1139–1147, 2013
Monday, September 18 Due: Review of Paper 2.

Lecture 4. [Slides] Topics:
  • The kernel trick
  • Gram matrix versus feature extraction: systems tradeoffs
  • Adaptive/data-dependent feature mappings
Wednesday, September 20 Paper Discussion 3. Ali Rahimi and Benjamin Recht. Random features for large-scale kernel machines. In Advances in neural information processing systems, pages 1177–1184, 2007
Monday, September 25 Due: Review of Paper 3.

Lecture 5. [Slides] Topics:
  • Online versus offline learning
  • Variance reduction
  • SVRG
  • Fast linear rates for convex objectives
Wednesday, September 27 Paper Discussion 4. Rie Johnson and Tong Zhang. Accelerating stochastic gradient descent using predictive variance reduction. In Advances in neural information processing systems, pages 315–323, 2013
Monday, October 2 Due: Review of Paper 4.

Lecture 6. [Slides] [Demo Notebook] Topics:
  • Metaparameter optimization
  • Assigning parameters from folklore
  • Random search over parameters
Wednesday, October 4 Paper Discussion 5. James Bergstra and Yoshua Bengio. Random search for hyper-parameter optimization. Journal of Machine Learning Research, 13(Feb): 281–305, 2012
Monday, October 9 Fall break. No lecture.
Wednesday, October 11 Due: Review of Paper 5.

Paper Discussion 6. Jasper Snoek, Hugo Larochelle, and Ryan P Adams. Practical bayesian optimization of machine learning algorithms. In Advances in neural information processing systems, pages 2951–2959, 2012
Monday, October 16 Due: Review of Paper 6.

Lecture 7. [Slides] Topics:
  • Non-convex stochastic gradient descent
  • Weakness of theoretical guarantees
  • One case where we can say something: stochastic power iteration
  • Deep learning as non-convex optimization
Wednesday, October 18 Paper Discussion 7. Diederik Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014
Monday, October 23 Due: Review of Paper 7.

Lecture 8. [Slides] Topics:
  • Major bottleneck for ML systems: parallelism
  • Asynchronous execution
  • Hogwild!
Wednesday, October 25 Paper Discussion 8. Feng Niu, Benjamin Recht, Christopher Re, and Stephen Wright. Hogwild: A lock-free approach to parallelizing stochastic gradient descent. In Advances in neural information processing systems, pages 693–701, 2011
Monday, October 30 Due: Review of Paper 8.

Lecture 9. [Slides] Topics:
  • Major bottleneck for ML systems: memory bandwidth and locality
  • Low precision computation
  • Vector computation
  • Scan orders
Wednesday, November 1 Paper Discussion 9. Suyog Gupta, Ankur Agrawal, Kailash Gopalakrishnan, and Pritish Narayanan. Deep learning with limited numerical precision. In International conference on machine learning, 2015
Monday, November 6 Due: Review of Paper 9.

Lecture 10. [Slides] Topics:
  • Algorithms other than SGD
  • What happens on the inference side?
  • Stochastic coordinate descent
  • Markov chain Monte Carlo and Gibbs sampling
  • Contrastive divergence
  • Derivative free optimization
Wednesday, November 8 Paper Discussion 10. Song Han, Huizi Mao, and William J Dally. Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding. ICLR, 2016
Monday, November 13 Due: Review of Paper 10.

Due: Final Project Proposal.

Lecture 11. [Slides] Topics:
  • Hardware for machine learning
  • The dominance of GPUs
  • Accelerators for machine learning
  • Will all computation become matrix multiply?
Wednesday, November 15 Paper Discussion 11. Norman P Jouppi, Cliff Young, Nishant Patil, David Patterson, Gaurav Agrawal, Raminder Bajwa, Sarah Bates, Suresh Bhatia, Nan Boden, Al Borchers, et al. In-datacenter performance analysis of a tensor processing unit. In Proceedings of the 44th Annual International Symposium on Computer Architecture, pages 1–12. ACM, 2017
Monday, November 20 Due: Review of Paper 11.

Lecture 12. [Slides] Topics:
  • Machine learning frameworks and cluster parallelism
  • TensorFlow
  • SciKit-Learn
  • PyTorch
  • Is Python the ML language of the future?
Wednesday, November 22 Thanksgiving break. No lecture.
Monday, November 27 Due: Abstract for Final Project.

Abstract swap and discussion.
Wednesday, November 29 Lecture 14. [Slides] Topics:
  • Sparsity
  • Structured matrices
Wednesday, December 6 Due: Final Project Report.