Course Overview

Large datasets pose difficulties across the machine learning pipeline. They are difficult to visualize, and it can be hard to determine what sorts of errors and biases may be present in them. They are computationally expensive to process, and the cost of learning is often hard to predict---for instance, an algorithm that runs quickly on a dataset that fits in memory may be exorbitantly expensive when the dataset is too large for memory. Large datasets may also display qualitatively different behavior in terms of which learning methods produce the most accurate predictions.

This course is intended to provide a student practical knowledge of, and experience with, the issues involving large datasets. Among the topics considered are: data cleaning, visualization, and pre-processing at scale; principles of parallel and distributed computing for machine learning; techniques for scalable deep learning; analysis of programs in terms of memory, disk usage, and (for parallel methods) communication complexity; and methods for low-latency inference. Students will gain experience with common large-scale computing libraries and infrastructure, including Apache Spark and TensorFlow.

Prerequisites

Students are required to have taken a CMU introductory machine learning course (10-301, 10-315, 10-601, 10-701, or 10-715). A strong background in programming will also be necessary; suggested prerequisites include 15-210, 15-214, or equivalent. Students are expected to be familiar with Python or learn it during the course.

Textbooks

There will be no required textbooks, though we may suggest additional reading in the schedule below.

Course Components

The requirements of this course consist of participating in lectures, homework assignments, and two exams. The grading breakdown is the following:

Exams

You are required to attend all exams. The exams will be given during class. Please plan your travel accordingly as we will not be able accommodate individual travel needs (e.g. by offering the exam early).

If you have an unavoidable conflict with an exam (e.g. an exam in another course), notify us by filling out the exam conflict form which will be released on Piazza a few weeks before the exam.

Homework

The homeworks will be divided into two components: programming and written. The programming assignments will ask you to implement ML algorithms from scratch; they emphasize understanding of real-world applications of ML, building end-to-end systems, and experimental design. The written assignments will focus on core concepts, “on-paper” implementations of classic learning algorithms, derivations, and understanding of theory.

Technologies

Piazza

We will use Piazza for class discussions. Please go to this Piazza website to join the course forum (note: you must use a cmu.edu email account to join). We strongly encourage students to post on this forum rather than emailing the course staff directly (this will be more efficient for both students and staff). Students should use Piazza to:

The course Academic Integrity Policy must be followed on the message boards at all times. Do not post or request homework solutions! Also, please be polite.

Gradescope

We use Gradescope to collect PDF submissions of open-ended questions on the homework (e.g. mathematical derivations, plots, short answers). The course staff will manually grade your submission, and you’ll receive personalized feedback explaining your final marks.

You will also submit your code for programming questions on the homework to Gradescope. After uploading your code, our grading scripts will autograde your assignment by running your program on a VM. This provides you with immediate feedback on the performance of your submission.

Regrade Requests

If you believe an error was made during manual grading, you’ll be able to submit a regrade request on Gradescope. For each homework, regrade requests will be open for only 1 week after the grades have been published. This is to encourage you to check the feedback you’ve received early!

Schedule (Subject to Change)

Date Lecture Resources Announcements
Data Pre-Processing and Visualization, Distributed Computing
Feb 1 Introduction
Feb 3 Distributed Computing, Spark
Feb 8 Visualization, PCA Tutorial on PCA
JL Theorem
Feb 10 Nonlinear Dimensionality Reduction t-SNE
Basics of Large-Scale / Distributed Machine Learning
Feb 15 Distributed Linear Regression, part I
Feb 17 Distributed Linear Regression, part II
Feb 22 Kernel Approximations
Feb 24 Guest Lecture Theodoros (Theo) Rekatsinas Scheduled at 11:00am instead of 8:20am
Mar 1 Logistic Regression, Hashing Hash kernels, I
Hash kernels, II
Mar 3 Randomized Algorithms Count-min sketch
LSH
Mar 8 Distributed Trees
Mar 10 TBA
Mar 15 Exam I
Mar 17 Cloud Computing & Services
Scalable Deep Learning: Training, Tuning, and Inference
Mar 22 Deep Learning, Autodiff Deep Learning, Ch. 6
TensorFlow Quickstart
Mar 25 DL Frameworks, Hardware
Mar 29 Large-Scale Optimization Optimization for Large-Scale ML
Mar 31 Optimization for DL Deep Learning, Ch. 8
Apr 5 No Class (CMU Break Day)
Apr 7 Parallel/Distributed DL
Apr 12 Hyperparameter Tuning blog1, blog2
Apr 14 Inference, Model Compression
Advanced Topics
Apr 19 Neural Architecture Search blog, RSWS
Apr 21 Guest Lecture
Apr 26 Productionizing Large-Scale ML
Apr 28 Federated Learning
May 3 Course Summary
May 5 Exam II

General Policies

Late Homework Policy

You receive 4 total grace days for use on any homework assignment. We will automatically keep a tally of these grace days for you; they will be applied greedily. No assignment will be accepted more than 1 days after the deadline without written permission from Daniel, Ameet, or Virginia. You may not use more than 1 grace day on any single assignment.

All homework submissions are electronic (see Technologies section below). As such, lateness will be determined by the latest timestamp of any part of your submission. For example, suppose the homework requires submissions to both Gradescope Written and Programming– if you submit your Written on time but your Programming 1 minute late, your entire homework will be penalized for the full 24-hour period.

Extensions

In general, we do not grant extensions on assignments. There are several exceptions: For any of the above situations, you may request an extension by emailing Daniel Bird (dpbird@andrew.cmu.edu). The email should be sent as soon as you are aware of the conflict and at least 5 days prior to the deadline. In the case of an emergency, no notice is needed.

Audit Policy

Official auditing of the course (i.e. taking the course for an “Audit” grade) is not permitted this semester.

Unofficial auditing of the course (i.e. watching the lectures online or attending them in person) is welcome and permitted without prior approval. Unofficial auditors will not be given access to course materials such as homework assignments and exams.

Pass/Fail Policy

Pass/Fail is only allowed in this class with written permission from Ameet or Virginia. The grade for the Pass cutoff will depend on your program. Be sure to check with your program / department as to whether you can count a Pass/Fail course towards your degree requirements.

Accommodations for Students with Disabilities

If you have a disability and have an accommodations letter from the Disability Resources office, I encourage you to discuss your accommodations and needs with me as early in the semester as possible. I will work with you to ensure that accommodations are provided as appropriate. If you suspect that you may have a disability and would benefit from accommodations but are not yet registered with the Office of Disability Resources, I encourage you to contact them at access@andrew.cmu.edu.

Academic Integrity Policies

Read this Carefully

Collaboration among Students

Previously Used Assignments

Some of the homework assignments used in this class may have been used in prior versions of this class, or in classes at other institutions, or elsewhere. Solutions to them may be, or may have been, available online, or from other people or sources. It is explicitly forbidden to use any such sources, or to consult people who have solved these problems before. It is explicitly forbidden to search for these problems or their solutions on the internet. You must solve the homework assignments completely on your own. We will be actively monitoring your compliance. Collaboration with other students who are currently taking the class is allowed, but only under the conditions stated above.

Policy Regarding “Found Code”

You are encouraged to read books and other instructional materials, both online and offline, to help you understand the concepts and algorithms taught in class. These materials may contain example code or pseudo code, which may help you better understand an algorithm or an implementation detail. However, when you implement your own solution to an assignment, you must put all materials aside, and write your code completely on your own, starting “from scratch”. Specifically, you may not use any code you found or came across. If you find or come across code that implements any part of your assignment, you must disclose this fact in your collaboration statement.

Duty to Protect One’s Work

Students are responsible for proactively protecting their work from copying and misuse by other students. If a student’s work is copied by another student, the original author is also considered to be at fault and in gross violation of the course policies. It does not matter whether the author allowed the work to be copied or was merely negligent in preventing it from being copied. When overlapping work is submitted by different students, both students will be punished.

To protect future students, do not post your solutions publicly, neither during the course nor afterwards.

Penalties for Violations of Course Policies

All violations (even first one) of course policies will always be reported to the university authorities (your Department Head, Associate Dean, Dean of Student Affairs, etc.) as an official Academic Integrity Violation and will carry severe penalties.
  1. The penalty for the first violation is a one-and-a-half letter grade reduction. For example, if your final letter grade for the course was to be an A-, it would become a C+.
  2. The penalty for the second violation is failure in the course, and can even lead to dismissal from the university.

Acknowledgments

This course is based in part on material developed by Heather Miller, William Cohen, Anthony Joseph, and Barnabas Poczos.

Previous courses: 10-605/10-805, Fall 2020; 10-405/10-605, Spring 2020.