Overview

Machine Learning is concerned with computer programs that automatically improve their performance through experience (e.g., programs that learn to recognize human faces, recommend music and movies, and drive autonomous robots). This course covers the theory and practical algorithms for machine learning from a variety of perspectives. We cover topics such as Bayesian networks, decision tree learning, support vector machines, statistical learning methods, unsupervised learning and reinforcement learning. The course covers theoretical concepts such as inductive bias, the PAC learning framework, Bayesian learning methods, and margin-based learning. Programming assignments include hands-on experiments with various learning algorithms. This course is designed to give a graduate-level student a thorough grounding in the methodologies, technologies, mathematics and algorithms currently needed by people who do research in machine learning.

Learning Objectives

After completing the course, students should be able to:

• Implement and analyze existing learning algorithms, including well-studied methods for classification, regression, structured prediction, clustering, and representation learning
• Integrate multiple facets of practical machine learning in a single system: data preprocessing, learning, regularization and model selection
• Describe the the formal properties of models and algorithms for learning and explain the practical implications of those results
• Compare and contrast different paradigms for learning (supervised, unsupervised, etc.)
• Design experiments to evaluate and compare different machine learning techniques on real-world problems
• Employ probability, statistics, calculus, linear algebra, and optimization in order to develop new predictive models or learning methods
• Given a description of a ML technique, analyze it to identify (1) the expressive power of the formalism; (2) the inductive bias implicit in the algorithm; (3) the size and complexity of the search space; (4) the computational properties of the algorithm: (5) any guarantees (or lack thereof) regarding termination, convergence, correctness, accuracy or generalization power.

Levels

10-301 and 10-601 are identical. Undergraduates must register for 10-301 and graduate students must register for 10-601. This course covers many similar topics to other introductory machine learning and A.I. courses, such as 10-315, 10-701, 15-281. Contact the instructor if you are concerned about which course is appropriate for you.

Prerequisites + Corequisites

Students entering the class are expected to have a pre-existing working knowledge of probability, linear algebra, statistics and algorithms, though the class has been designed to allow students with a strong numerate background to catch up and fully participate. In addition, recitation sessions will be held to review some basic concepts.

1. You need to have, before starting this course, significant experience programming in a general programming language. Specifically, you need to have written from scratch programs consisting of several hundred lines of code. For undergraduate students, this will be satisfied for example by having passed 15-122 (Principles of Imperative Computation) with a grade of ‘C’ or higher, or comparable courses or experience elsewhere. Note: For each programming assignment, we will allow you to pick between Python, C++, and Java.
2. You need to have, before starting this course, basic familiarity with probability and statistics, as can be achieved at CMU by having passed 36-217 (Probability Theory and Random Processes) or 36-225 (Introduction to Probability and Statistics I), or 15-359, or 21-325, or comparable courses elsewhere, with a grade of ‘C’ or higher.
3. You need to have, before starting this course, college-level maturity in discrete mathematics, as can be achieved at CMU by having passed 21-127 (Concepts of Mathematics) or 15-151 (Mathematical Foundations of Computer Science), or comparable courses elsewhere, with a grade of ‘C’ or higher.

You must strictly adhere to these pre-requisites! Even if CMU’s registration system does not prevent you from registering for this course, it is still your responsibility to make sure you have all of these prerequisites before you register.

Notably missing in this prerequisite list is any linear algebra course. Linear algebra is indeed a central piece to this machine learning course. Given the lack of a linear algebra prerequisite, we will provide the necessary resources and instruction for linear algebra. That being said, if you have never been exposed to matricies and vectors in any context, please contact the instructor to discuss how to best meet your linear algebra needs.

Please see the instructor if you are unsure whether your background is suitable for the course.

Office Hours

See office hours on the calendar below.

When appropriate, this course uses the CMU OHQueue tool as a queueing system for office hours.

Feel free to contact the course staff via a private post on Piazza to request office hours by appointment. We'll do our best to accommodate these requests. We also occasionally open 15-minute appointment slots on the course calendar. These appointment slots may be secured via the following link. Please be courteous to other students when selecting these slots, e.g. don't select more than one slot per day. Link: OH Appointment Slots

Schedule (subject to change)

Textbooks:

Daumé III, Hal. A Course in Machine Learning, available online, (optional)

Bishop, Christopher. Pattern Recognition and Machine Learning, available online, (optional)

Murphy, Kevin P. Machine Learning: A Probabilistic Perspective, available online, (optional)

Goodfellow, Ian, Yoshua Bengio, Aaron Courville. Deep Learning, available online, (optional)

Mitchell, Tom. Machine Learning, select chapters available online, (optional)

Dates	Topic	Reading / Demo	Slides / Notes
8/31 Mon	1: Introduction to ML	notation.pdf	pptx (inked) pdf (inked)
9/1 Wed	2: Decision Trees	Daumé 1	pptx (inked) pdf (inked)
9/7 Mon	No class: Labor Day
9/9 Wed	3: Decision Trees	Daumé 2, Entropy, Cross-Entropy video, A. Géron	pptx (inked) pdf (inked)
9/14 Mon	4: Decision Trees and Nearest Neighbor	Daumé 3	pptx (inked) pdf (inked)
9/16 Wed	5: Nearest Neighbor and Model Selection		pptx (inked) pdf (inked)
9/21 Mon	6: Linear Regression	Murphy 7.1-7.3	pptx (inked) pdf (inked)
9/23 Wed	7: Optimization		pptx (inked) pdf (inked)
9/28 Mon	8: Logistic Regression	Murphy 1.4.6, 8.1-8.3 lec8.ipynb	pptx (inked) pdf (inked)
9/30 Wed	9: Logistic Regression		pptx (inked) pdf (inked)
10/2 Fri	10: Logistic Regression and Feature Engineering		pptx (inked) pdf (inked)
10/5 Mon	MIDTERM EXAM 1, in-class
10/7 Wed	11: Feature Engineering and Regularization	Goodfellow, et al, Ch. 7.1, 7.8	pptx (inked) pdf (inked)
10/12 Mon	12: Regularization and Neural Networks	lec12.ipynb	pptx (inked) pdf (inked)
10/14 Wed	13: Neural Networks	Goodfellow, et al, Ch. 6	pptx (inked) pdf (inked)
10/19 Mon	14: Neural Networks	Goodfellow, et al, Ch. 9	pptx (inked) pdf (inked)
10/21 Wed	15: Learning Theory	A Few Useful Things to Know about Machine Learning. Pedro Domingos (2012).	pptx (inked) pdf (inked)
10/26 Mon	16: Learning Theory	Generalization Abilities: Sample Complexity Results. Nina Balcan (2015). Lecture notes.	PAC Learning: Theorem 1 pptx (inked) pdf (inked)
10/28 Wed	17: MLE/MAP	Estimating Probabilities: MLE and MAP. Tom Mitchell (2018, draft)	pptx (inked) pdf (inked)
11/2 Mon	18: Generative Models and Naive Bayes	Generative and Discriminative Classifiers. Tom Mitchell (2020, draft).	pptx (inked) pdf (inked) SPAM handout (sol)
11/4 Wed	19: Bayes Nets	15-281 Probability Reference Sheet	pptx (inked) pdf (inked)
11/6 Fri	20: HMMs	A Tutorial on HMMs. Rabiner (1989). [Only pages 257 - 266] You may need to authenticate here first: CMU Libraries IEEE	pptx (inked) pdf (inked)
11/9 Mon	MIDTERM EXAM 2, in-class
11/11 Wed	21: HMMs		pptx (inked) pdf (inked)
11/16 Mon	22: MDPs	Reinforcement Learning: A Survey. Kaelbling, et al (1996).	pptx (inked) pdf (inked)
11/18 Wed	23: MDPs		pptx (inked) pdf (inked)
11/20 Fri	24: Reinforcement Learning	[Additional] Playing Atari with Deep Reinforcement Learning. Mnih, et al (2013).	pptx (inked) pdf (inked)
11/23 Mon	Recitation
11/25 Wed	No class: Thanksgiving
11/30 Mon	25: Support Vector Machines	Bishop 7.1 Daumé 11	pptx (inked) pdf (inked)
12/2 Wed	26: Support Vector Machines		pptx (inked) pdf (inked)
12/7 Mon	Clustering	Bishop 9.1	pptx (inked) pdf (inked)
12/9 Wed	Dimensionality Reduction	A Tutorial on Principal Component Analysis. Jonathon Shlens (2014).	pptx (inked) pdf (inked)
12/14 Mon	FINAL EXAM	8:30-11:30 am or 1:00-4:00 pm

Recitation

There will be three recitation sections on Fridays: 8:00 am, 11:40 am, and 3:20 pm. You may attend any one of these sections.
Recitation will take place live using Zoom. See Canvas for Zoom links. Zoom sessions for recitation will not be recorded.

Recitation attendance is recommended to help solidfy weekly course topics. That being said, the recitation materials published below are required content and are in-scope for midterm and final exams.

Dates	Recitation	Handout	Code/Demo
9/4 Fri	Recitation 1	recitation1.pdf (solutions)
9/11 Fri	Recitation 2	recitation2.pdf (solutions)	Jupyter Notebook, recitation2.py
9/18 Fri	Recitation 3		Numpy Notebook, Logging Notebook, Workflow and Debugging
9/25 Fri	Recitation 4	recitation4.pdf (solutions)
10/2 Fri	Lecture
10/9 Fri	Recitation 5	recitation5.pdf (solutions)
10/16 Fri	Recitation 6	recitation6.pdf (solutions)
10/30 Fri	Recitation 7	recitation7.pdf (solutions)
11/13 Fri	Recitation 8	recitation8.pdf (solutions)
11/23 Fri	Recitation 9	recitation9.pdf (solutions)
12/4 Fri	Recitation 10	recitation10.pdf (solutions)

Exams

The course includes two midterm exams and a final exam. The midterms will both take place during your lecture timeslot on Monday, Oct. 5 and Monday, Nov. 9. The final exam is on Monday, Dec. 14, 8:30-11:30 am or 1:00-4:00 pm. Plan any travel around exams, as exams cannot be rescheduled.

Assignments

There will be approximately nine homework assignments that will have some combination of written and programming components. Written components will involve working through algorithms presented in the class, deriving and proving mathematical results, and critically analyzing material presented in class. Programming components will involve writing code in Python, C++, or Java to implement various algorithms.

For any assignments that aren't released yet, the dates below are tentative and subject to change.

Policies

Grading

Grades will be collected and reported in Canvas. Please let us know if you believe there to be an error the grade reported in Canvas.

Final scores will be composed of:

• 15% Midterm exam 1
• 15% Midterm exam 2
• 15% Final exam
• 50% Homework assignments
• 5% Participation

Participation Grades

Participation will be based on the percentage of in-class polling questions answered:

• 5% for 80% or greater poll participation
• 3% for 70%
• 1% for 60%

Correctness of in-class polling responses will not be taken into account for participation grades.

It is against the course academic integrity policy to answer in-class polls when you are not present in lecture. Violations of this policy will be reported as an academic integrity violation. Information about academic integrity at CMU may be found at https://www.cmu.edu/academic-integrity.

There will be a few other means to collect participation points; stay tuned to Piazza for more details.

Final Grade

This class is not curved. However, we convert final course scores to letter grades based on grade boundaries that are determined at the end of the semester. What follows is a rough guide to how course grades will be established, not a precise formula — we will fine-tune cutoffs and other details as we see fit after the end of the course. This is meant to help you set expectations and take action if your trajectory in the class does not take you to the grade you are hoping for. So, here's a rough, very rough heuristics about the correlation between final grades and total scores:

• A: above 90%
• B: 80-90%
• C: 70-80%
• D: 60-70%

Grades for graduate students will be broken down further with +/- distinctions. See CMU grading polices for more information.

The above heuristic assumes that the makeup of a student's grade is not wildly anomalous: exceptionally low overall scores on exams, programming assignments, or written assignments will be treated on a case-by-case basis.

Precise grade cutoffs will not be discussed at any point during or after the semester. For students very close to grade boundaries, instructors may, at their discretion, consider participation in lecture and recitation, exam performance, and overall grade trends when assigning the final grade.

Late Policy

Homework assignments:

• 6 slip days across all assignments
• Use up to two per assignment
• Slip days are counted by the granularity of days. For example, if you turn in your assignment 30 minutes after the deadline, that will count as one full slip day.
• You may use these at your discretion, but they are intended for minor illness and other disruptive events outside of your control, and not for poor time management
• You are responsible to keep track of your own slip days. Gradescope will not enforce the total number of slip days
• Homework submitted after these two slip days or submitted by a student without any slip days remaining will be given a score of 0.

Aside from this, there will be no extensions on assignments in general. If you think you really really need an extension on a particular assignment, contact the instructor as soon as possible and before the deadline. Please be aware that extensions are entirely discretionary and will be granted only in exceptional circumstances outside of your control (e.g., due to severe illness or major personal/family emergencies, but not for competitions, club-related events or interviews). The instructors will require confirmation from University Health Services or your academic advisor, as appropriate.
Nearly all situations that make you run late on an assignment homework can be avoided with proper planning — often just starting early. Here are some examples:

• I have so many deadlines this week: you know your deadlines ahead of time — plan accordingly.
• It's a minute before the deadline and the network is down: you always have multiple submissions - it's not a good idea to wait for the deadline for your first submission.
• My computer crashed and I lost everything: Use Dropbox or similar to do real-time backup - recover your files onto AFS and finish your homework from a cluster machine.
• My fraternity/sorority/club has that big event that is taking all my time: Schedule your extra-curricular activities around your classes, not vice versa.

Collaboration Policy

We encourage you to discuss course content and assignments with your classmates. However, these discussions must be kept at a conceptual level only.

• You may NOT view, share, or communicate about any artifact that will be submitted as part of an assignment. Example artifacts include, but are not limited to: code, pseudocode, diagrams, and text.
• You may look at another student's code output and discuss it at a conceptual level, as long as it is not output that appears directly in the homework submission.
• You may look at another student's code error messages and discuss what the error means at a conceptual level. However, you may NOT give specific instructions to fix the error.
• All work that you present must be your own.
• Using any external sources of code or algorithms in any way must have approval from the instructor before submitting the work. For example, you must get instructor approval before using an algorithm you found online for implementing a heuristic function in a programming assignment.

Violations of these policies will be reported as an academic integrity violation. Information about academic integrity at CMU may be found at https://www.cmu.edu/academic-integrity. Please contact the instructor if you ever have any questions regarding academic integrity or these collaboration policies.

Accommodations for Students with Disabilities

If you have a disability and have an accommodations letter from the Disability Resources office, we encourage you to discuss your accommodations and needs with us as early in the semester as possible. We 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, we encourage you to visit their website.

Statement of Support for Students’ Health & Well-being

Take care of yourself. Do your best to maintain a healthy lifestyle this semester by eating well, exercising, getting enough sleep, and taking some time to relax. This will help you achieve your goals and cope with stress.
All of us benefit from support during times of struggle. There are many helpful resources available on campus and an important part of the college experience is learning how to ask for help. Asking for support sooner rather than later is almost always helpful.
If you or anyone you know experiences any academic stress, difficult life events, or feelings like anxiety or depression, we strongly encourage you to seek support. Counseling and Psychological Services (CaPS) is here to help: call 412-268-2922 and visit their website at http://www.cmu.edu/counseling/. Consider reaching out to a friend, faculty or family member you trust for help getting connected to the support that can help.
If you have questions about this or your coursework, please let us know. Thank you, and have a great semester.

Research to Improve the Course

For this class, we are conducting research on teaching and learning. This research will involve some student work. You will not be asked to do anything above and beyond the normal learning activities and assignments that are part of this course. You are free not to participate in this research, and your participation will have no influence on your grade for this course or your academic career at CMU. If you do not wish to participate, please send an email to Chad Hershock (hershock@andrew.cmu.edu). Participants will not receive any compensation. The data collected as part of this research will include student grades. All analyses of data from participants’ coursework will be conducted after the course is over and final grades are submitted. The Eberly Center may provide support on this research project regarding data analysis and interpretation. The Eberly Center for Teaching Excellence and Educational Innovation is located on the CMU-Pittsburgh Campus and its mission is to support the professional development of all CMU instructors regarding teaching and learning. To minimize the risk of breach of confidentiality, the Eberly Center will never have access to data from this course containing your personal identifiers. All data will be analyzed in de-identified form and presented in the aggregate, without any personal identifiers. If you have questions pertaining to your rights as a research participant, or to report concerns to this study, please contact Chad Hershock (hershock@andrew.cmu.edu).