Syllabus: CS/DSA 5703: Machine Learning Practice (Fall 2022)

Topics

• Representing information and preparing data for use with ML methods
• Classifiers and feature importance, including K-nearest neighbors, logistic regression, support vector machines
• Decision trees: ensemble methods, random forests, and boosting
• Regression and combating over-fitting: ridge regression, Tikhonov regression, lasso, elastic nets, polynomial regression, support vector regression
• Nonlinear dimensionality reduction: kernel PCA, local linear embedding, ISOmap, multidimensional scaling
• Semi-supervised learning: label spreading, label propagation
• Unsupervised learning: mixture models
• Evaluation in ML: metrics, cross-validation, statistics, addressing the multiple comparisons problem

Teaching Approach

Over the course of the semester, we will talk about a wide range of algorithms and practice their use with various data-sets. Students will be expected to apply this knowledge by working with existing Python code and developing their own. The homework assignments that we will assign are intended to exercise all of these skills, often requiring extrapolation of the base knowledge. This approach is intended to be challenging, but you are not on your own. The instructor and the teaching assistant are here to answer questions, discuss solutions and offer general guidance. We expect that everyone will be proactive when they are stuck on something.

General Information

• There are two versions of the course:
  • In person: CS 5703-01. Carson 441. T/Th 9-10:15
  • All others: Online/Asynchronous. Lectures will be delivered using pre-recorded videos.

• Prerequisites:
  • We rely on a background that includes programming (including object-oriented programming), regression and hypothesis testing

  • Specific prerequisites: Artificial Intelligence (CS 4013/5013) OR Data Mining (CS 5593) OR Intelligent Data Analytics (ISE/DSA 5103) OR permission of the instructor (presuming similar background).

• NOTE: If you have already taken the CS Machine Learning course (CS 4033/5033), then you really should be taking one of our higher-division ML courses.

• Reading Materials:
  • Aurélien Géron (2020) Hands-On Machine Learning with Scikit-Learn, Keras and TensorFlow (Concepts, Tools, and Techniques to Build Intelligent Systems), 2nd edition, ISBN-13: 978-1492032649, O'Reilly Media
  • Various papers and other network resources.

• Our new reality: as with other recent semesters, it is going to be tough for everyone this semester. We are dedicated to helping you all have a meaningful and quality experience in this class. We ask that you talk to us any time that you are having difficulty. However, please work to meet us part way: keep up with the readings and try writing bits of code to help clarify ideas. And - don't be afraid to fail in some of the things that you do for this class. Often, we learn a lot more failing than when we are succeeding.

• Other key materials:
  • Class schedule: https://symbiotic-computing.org/fagg_html/classes/mlp/schedule.html

  • Python 3 API Reference

  • Pandas data analysis package

  • Matplotlib

  • Numpy

  • Scikit-Learn

  • Jupyter Lab: https://jupyterlab.readthedocs.io/en/stable/

• Course web page: https://symbiotic-computing.org/fagg_html/classes/mlp

• We will also be making heavy use of Canvas

• Instructor: Dr. Andrew H. Fagg
  • Office: DEH 243
  • Office hours will be held on GatherTown. See the Canvas syllabus for the invitation link
  • Phone: 325-8606 (note that I do not receive voicemail from this phone)
  • Homepage: https://symbiotic-computing.org/fagg_html
  • Email: andrewhfagg--gmail.com
  • Office hours: see the office hours web page https://symbiotic-computing.org/fagg_html/office.html

• Teaching Assistant: Aditya Rohan Singh
  • Office hours will be held on GatherTown
  • Email: Aditya.Rohan.Singh-1 -- ou.edu
  • Office hours: see the office hours web page https://symbiotic-computing.org/fagg_html/office.html

Course Policies

• Lectures:
  • In-Person: Lectures are T/Th 9-10:15 in Carson 441.

  • Online: Lecture material will be distributed using videos available through the Canvas page. These videos will be released at the beginning of each week. Students should plan to view the videos within the same week.

    Note that the videos are distributed through YouTube playlists (so, each week's lectures are split into a large number of small videos).

• Class Web Page: Most of the material that you will need can be found on the class web page located at: https://symbiotic-computing.org/fagg_html/classes/mlp

• Canvas: This class will also use Canvas, located at: http://canvas.ou.edu

  Login with your 4+4 (typically the first four letters of your last name followed by the last four digits of your student number), using your standard OU password. If you have difficulty logging in, call 325-HELP. This software provides a number of useful features, including a list of assignments and announcements, an electronic mailing list, newsgroups, and a grade book.

  I may update the main web site and the Canvas page several times a week. When I update the site in any significant way, I will post an announcement on Canvas telling you what has been added and where it is located. You are responsible for things posted on the site within 48 hours of the post.

• Class Communication:
  • I will hold regular, weekly office hours using GatherTown, as will the teaching assistant. My office hours can also be attended in person. Office hours are listed on my office hours page

  • Other than the public office hours, the Slack page or the discussion group on Canvas should be the primary method of communication. This allows everyone in the class to benefit from the answers to your questions, and provides students with more timely answers.

  • Matters of personal interest should be directed to email instead of to the newsgroup, e.g. informing me of an extended illness.

  • Announcements will be posted to the Canvas announcement board.

  • It is your responsibility to have Canvas configured so that you receive these messages in a timely fashion. Note that Canvas can be configured so that it will forward messages, discussion posts and announcements directly to your email address.

  • Homework assignments and associated code will be posted to Google Drive. The link will be posted in Canvas

• Final Exam: There is no final exam for this course.

• Proper Academic Conduct:
  • Homework assignments must be your own work.

  • While the net may be used as a reference, downloading code that solves all or a substantial portion of any assignment from the net is prohibited. Note that programs will be checked by software designed to detect improper copying. This software is extremely effective and has withstood repeated reviews by the campus judicial processes.

  • Upon the first documented occurrence of inappropriate collaborative work or downloading of solutions, I will report the academic misconduct to the Campus Judicial Coordinator. The procedure to be followed is documented in the University of Oklahoma Academic Misconduct Code (http://integrity.ou.edu).

  • Examples
    1. You and a friend in the class discuss the details of a project, including drawing ML algorithms on a whiteboard. Is this a case of academic misconduct? No

    2. You and a friend discuss the details of a project and share snippets of code that illustrate how to use Python or Scikit-Learn. Is this a case of academic misconduct? No

    3. You and a friend discuss the details of a project and share snippets of code that solve parts of the project. Is this a case of academic misconduct? Yes

    4. In the presence of the instructor or TA, you share snippets of code that solve parts of the project in order to receive advice; other students are present for the conversation. Is this a case of academic misconduct? No, but we may decide to move the conversation out of sight of the other students.

    5. A friend who took the same class in a prior semester offers to provide project solutions. Is it academic misconduct to accept this code? Yes

    6. You store your homework code on a service such as Github. The code is searchable and downloadable by anyone. Is this a case of academic misconduct? Yes

    7. You store your homework code on a service such as Github. The code is not searchable, but you share a link with someone else in the class. Is this a case of academic misconduct? Yes

    8. You store your homework code on a service such as Github. The code is not accessible by others. Is this a case of academic misconduct? No

    9. You post homework problems to a service such as Chegg. Is it misconduct to receive help on these problems or to provide answers to these problems to the service? Yes

    10. You discuss the solution of problems similar to what can be found in a homework assignment. Is this a case of academic misconduct? No

• Incompletes: The grade of "I" is intended for the rare circumstance when a student who has been successful in a class has an unexpected event occur shortly before the end of the class. I will not consider giving a student a grade of "I" unless the following three conditions have been met.
  • It is within two weeks of the end of the semester.
  • The student has a grade of C or better in the class.
  • The reason that the student cannot complete the class is properly documented and compelling.

• Accommodation of Disabilities: The University of Oklahoma is committed to providing reasonable accommodation for all students with disabilities. Students with disabilities who require accommodations in this course are requested to speak with the professor as early in the semester as possible. Students with disabilities must be registered with the Accessibility and Disability Resource Center prior to receiving accommodations in this course.

Grades

• Final course grades will be based on performance on 13 homework assignments given over the course of the semester. The top 12 of the 13 homework grades will be used to compute the final grade.

• Homework assignments are due on the day indicated on the course schedule at 11:59pm. Each assignment must be turned in the following way:
  • To Gradescope: turn in the notebook file itself (.ipynb).

• Late policy:
  • Homework assignments will be accepted one day late with a 10% penalty or two days late with a 20% penalty.
  • Accommodations can be negotiated should someone need more than one excused assignment.

General Grade Issues

• Grade questions: If you have a question about grading (including assessment of points), you may address these during office hours or in email. Note that if you are asking me to reconsider a grade, then I will likely re-examine the entire assignment. You have one week from the point that you receive feedback to address grading questions.

• Canvas Grade Summary: Canvas has a grade book that is used to store the raw data that is used to calculate your course grade. It is the responsibility of each student in this class to check their grades on Canvas after each assignment is graded. If an error is found, please bring it to my attention.

Inclusive Learning Environment

Food Pantry

Land Acknowledgment

Course Evaluations

Adjustments for Pregnancy/Childbirth Related Issues

Title IX Resources

Registration and Withdrawal

Emergency Protocol

• Severe Weather: If you receive an OU Alert to seek refuge or hear a tornado siren that signals severe weather 1. LOOK for severe weather refuge location maps located inside most OU buildings near the entrances 2. SEEK refuge inside a building. Do not leave one building to seek shelter in another building that you deem safer. If outside, get into the nearest building. 3. GO to the building's severe weather refuge location. If you do not know where that is, go to the lowest level possible and seek refuge in an innermost room. Avoid outside doors and windows. 4. GET IN, GET DOWN, COVER UP. 5. WAIT for official notice to resume normal activities.

  Links: Severe Weather Refuge Areas, Severe Weather Preparedness

• Armed Subject/Campus Intruder: If you receive an OU Alert to shelter-in-place due to an active shooter or armed intruder situation or you hear what you perceive to be gunshots: 1. GET OUT: If you believe you can get out of the area WITHOUT encountering the armed individual, move quickly towards the nearest building exit, move away from the building, and call 911. 2. HIDE OUT: If you cannot flee, move to an area that can be locked or barricaded, turn off lights, silence devices, spread out, and formulate a plan of attack if the shooter enters the room. 3. TAKE OUT: As a last resort fight to defend yourself.

  Links: OU Emergency Preparedness, Responding to Gunshots

• Fire Alarm/General Emergency: If you receive an OU Alert that there is danger inside or near the building, or the fire alarm inside the building activates: 1. LEAVE the building. Do not use the elevators. 2. KNOW at least two building exits 3. ASSIST those that may need help 4. PROCEED to the emergency assembly area 5. ONCE safely outside, NOTIFY first responders of anyone that may still be inside building due to mobility issues. 6. WAIT for official notice before attempting to re-enter the building.

  Links: OU Fire Safety on Campus

Copyright notice: Many of the materials created for this course are the intellectual property of Andrew H. Fagg. This includes, but is not limited to, the syllabus, lectures and course notes. Except to the extent not protected by copyright law, any sale of such materials requires the permission of the instructor.