Schedule

Teachers

Name	Office Hours	Email (please ask most questions in person)
Gillett, John (Teaching Faculty)	Tu 10:55-11:45 and 2:25-2:55, Th 11:55-12:45 in Morgridge 4508	jgillett@wisc.edu
Baiheng Chen (TA)	last week only: Fr 5/1/26 9:30-11:30 via Zoom, meeting ID 891 215 5818 (before last week it was Th 2:25-4:05 in Morgridge 5579)	bchen342@wisc.edu

Class Times

LEC 451-001, -002: TuTh 1:00-2:15 in Morgridge B2590

Course Description
Introduction to machine learning for pattern classification, regression analysis, clustering, and dimensionality reduction. For each category, fundamental algorithms, as well as selections of contemporary, current state-of-the-art algorithms, are being discussed. The evaluation of machine learning models using statistical methods is a particular focus of this course. Statistical pattern classification approaches, including maximum likelihood estimation and Bayesian decision theory, are compared and contrasted to algorithmic and nonparametric approaches. While fundamental mathematical concepts underlying machine learning and pattern classification algorithms are being taught, the practical use of machine learning algorithms using open source libraries from the Python programming ecosystem will be of equal focus in this course.

Learning Outcomes
After completing this course, a successful student will:

• Understand machine learning concepts:
  • Supervised learning:
    • Support vector machine
    • Linear regression
    • Logistic regression
    • Kernel regression
    • Decision tree classification and regression
    • k-NN classification and regression
  • Unsupervised learning:
    • Density estimation
    • Clustering
    • Dimensionality reduction
    • Outlier detection
  • Ensemble learning:
    • Boosting
    • Bagging
    • Gradient boosting
  • Basic practices:
    • Optimizing a cost function with gradient descent
    • Feature engineering
    • Splitting data into training, validataion, and test sets
    • Underfitting, overfitting, bias, variance, regularization
    • Hyperparameter tuning
    • Cross-validation
    • Algorithmic efficiency and model selection
    • Assessing model performance
• Use Python and its pandas, NumPy, Matplotlib, and SciKit-Learn libraries to apply machine learning to real data:
  • Use Jupyter Lab to write reports integrating text, data, code, and its output.
  • Manipulate data to create lists, arrays, and data frames.
  • Produce graphics.
  • Write functions, conditional expressions, and loops.
  • Extract data from text via regular expressions.

Requisites
MATH 320, 321, 340, 341, graduate/professional standing, or member of the Statistics Visiting International Scholars program

Designations and Attributes
Level: Advanced
Breadth: Natural Science
L&S Credit Type: Counts as Liberal Arts and Science credit (L&S)

Credit Information
This course is 3-credits. The class meets for two 75-minute in-person lectures each week and carries the expectation that students will work on course learning activities (readings, quizzes, homeworks, studying, etc.) for about 3 hours out of the classroom for every lecture period.

Instructional Mode
Classroom instruction

• See this schedule.
• There are nine machine learning topics in the schedule labeled "01 Introduction" through "09 Unsupervised Learning". There are also eight Python topics labeled "01 Python as a Calculator" through "08 Conditional Expressions". (The Python topics are colored orange to distinguish them.) For each topic, there are links to these types of course materials:
  • There are lecture notes that also serve as reference sheets consisting of ".pdf" files.
  • There are ".ipynb" (Jupyter Notebook) code example files, along with their compiled ".html" versions.
  • There are classroom lectures guided by the notes and code examples.
  • For each Python topic, there is an online quiz to provide low-stakes practice. You may repeat it many times. Your best score before its deadline counts. Each quiz includes 5 to 18 questions. Most questions can be answered by writing a single line of code.
• There are about 12 in-class group quizzes on machine learning.
• There are five homework assignments, each requiring up to about 100 lines of code.
• There is an in-class midterm exam and an in-class final exam.
• There is a semeseter project to be completed in groups of 5 students.
• There are two ways to get help:
  • Ask questions in class.
  • Ask questions in teacher and TA office hours.

Regular and Substantive Student-Instructor Interaction
The regular and substantive student-instructor interaction requirement is met through in-person lectures, regular weekly office hours, and homework, exam, and project feedback from teacher and TA.

Online materials
Course materials including online quizzes are posted in the schedule (linked above). Canvas is used to collect homework and as a gradebook.

Textbook

There is no required textbook.

Optional Reference Books and Videos

The Hundred-Page Machine Learning Book by Andriy Burkov

Machine Learning with PyTorch and Scikit-Learn (paper; eBook only by Sebastian Raschka, Yuxi Liu, and Vahid Mirjalili)

Python Machine Learning 3e by Sebastian Raschka and Vahid Mirjalili

Introduction to Machine Learning videos and notes by Sebastian Rashka

Think Python 2e by Allen Downey

Python for Beginners videos by Microsoft

Computing
A laptop with the free program Anaconda installed is required in class.

In case of trouble with Anaconda, an alternative is Google Colab (FAQ).

In case of trouble with your laptop, UW's InfoLabs offer loaner laptops and computer labs that have Anaconda on their Windows machines.

Grades

Grades are based on ≈ 400 points allocated to these tasks:

•   ≈ 50 points for ≈ 12 in-class group quizzes that follow group practice:
  • Groups of 6 students are randomly assigned to tables, with new groups for each quiz.
  • Groups solve practice questions while trying to get all members to understand.
  • Students take an individual quiz.
  • • ORIGINAL POLICY:
      The group average score (or 0 for non-participating students) is entered for each student in Canvas. This incentivizes immediate collaborative learning during class.
    • PRESENT POLICY:
      Students may work in groups. Each student receives his or her individual score.
  • Canvas will drop each student's 2 lowest group quizzes.
  • Group quizzes have advantages and disadvantages. Students may opt out because I will calculate a each student's course average with and without group quizzes, and use the higher average in assigning a letter grade. (I do not know how to implement this in Canvas, but I will calculate grades with this scheme after the midterm exam and post an anonymized gradebook. I will calculate the scheme again after the final exam.)
•   75 for 9 online Python quizzes:
  • The first, Q00, asks about your background.
  • The other eight, Q01-Q08, are designed to help you learn the basics of Python so that you can use it for machine learning. You may repeat Python quizzes. Your best score before the deadline counts.
• 50 for five machine learning software homework exercises, HW01-HW05. Your homework solutions should be submitted via file upload to Canvas.
• 100 for a midterm exam March 12 in class.
  
• 75 for a final exam May 3 7:45-9:45 a.m.
  
• 50 for a group machine learning project including a presentation:
  • Form a group (0 points) of 4-5 students (or be randomly assigned to a group).
  • Proposal (4 points) including code to read data, descriptions of the variables and machine learning methods you will use.
  • Meet with teacher and TA to discuss proposal (3 points).
  • Presentation (25 points) in class (for in-person courses) or via video (for online courses), evaluated via peer feedback.
  • Watch presentations (8 points spread over two or three days) and give peer feedback.
  • Report (10 points) of no more than 600 words with supporting graphics; and give peer feedback on two reports.

We will assign grades according to the percentage scale, A = [92,100], AB = [88,92), B = [82,88), BC = [78,82), C = [70,78), D = [60,70), F = [0,60) (92% of points ⇒ A); and according to the percentile scale A = 60, AB = 40, B = 20, BC = 8, C = 4, D = 2, F = 0 (which yields 40% A grades, 20% AB, 20% B, 12% BC, 4% C, 2% D, 2% F). Your grade will be the higher of these two grades.

If you anticipate religious or other conflicts with course requirements, or if you require accommodation due to disability, you must notify me during the first three weeks of class. You may not make up missed quizzes, homework, or exams, except in the rare case of a documented, serious problem beyond your control. Regarding late work:

• Group quizzes do not allow late submission. (But see the first bullet under Grades, above: Canvas drops each student's 2 lowest group quizzes and students may opt out.)
• Python quizzes are auto-graded by my web-form program. I will copy their scores to Canvas one day after their deadline; that is, you get one free late day for each quiz. (I recommend solving the quizzes on time, as if you do not, you are falling behind.)
• Canvas allows late submission of HW01-HW05 with a 1% per hour penalty. For example:
  • 1 second to 1 hour late gets a 1% penalty.
  • 12 hours late gets a 12% penalty.
  • 50 hours late gets a 50% penalty.
  (Note that this 1% per hour penalty will not excuse late project components, as late project work would delay presentations and peer feedback.)
• Exam and project deadlines are real, with no late submission permitted.

How to Succeed in This Course
The successful student will attend lectures; submit quizzes, homeworks, and project components on time; attend exams well-prepared; and ask questions when things are unclear.

Academic Policies and Statements

• Academic Calendar and Religious Observances
• Academic Integrity

  The members of the faculty of the Department of Statistics at UW-Madison uphold the highest ethical standards of teaching, data, and research. They expect their students to uphold the same standards of ethical conduct. Standards of ethical conduct in data analysis and data privacy are detailed on the ASA website (https://www.amstat.org/your-career/ethical-guidelines-for-statistical-practice), and include:

  • Use methodology and data that are relevant and appropriate; without favoritism or prejudice; and in a manner intended to produce valid, interpretable, and reproducible results.
  • Be candid about any known or suspected limitations, defects, or biases in the data that may affect the integrity or reliability of the analysis. Obviously, never modify or falsify data.
  • Protect the privacy and confidentiality of research subjects and data concerning them, whether obtained from the subjects directly, other persons, or existing records.
  By registering for this course, you are implicitly agreeing to conduct yourself with the utmost integrity throughout the semester.
• Accommodations for Students with Disabilities
• Course Evaluations
• Diversity and Inclusion
• Mental Health and Well-Being
• Privacy of Student Records and the Use of Audio Recorded Lectures
• Remote Proctoring with Honorlock (if applicable)
• Students' Rules, Rights and Responsibilities
• Teaching & Learning Data Transparency
• Complaints

  If you have a complaint about a TA or course instructor, you should feel free to discuss the matter directly with the TA or instructor. If the complaint is about the TA and you do not feel comfortable discussing it with him or her, you should discuss it with the course instructor. Complaints about mistakes in grading should be resolved with the instructor in the great majority of cases. If the complaint is about the instructor (other than ordinary grading questions) and you do not feel comfortable discussing it with him or her, contact the Director for Undergraduate Studies, Professor Jessi Cisewski-Kehe (jjkehe@wisc.edu) / Director of Graduate Studies, Professor Cecile Ane (cecile.ane@wisc.edu).

  If your complaint concerns sexual harassment, please see campus resources listed at https://compliance.wisc.edu/titleix/. In particular, there are a number of options to speak to someone confidentially.

  If you have concerns about climate or bias in this class, or if you wish to report an incident of bias or hate that has occurred in class, you may contact the Chair of the Statistics Department Climate & Diversity Committee, Professor Karl Rohe (karl.rohe@wisc.edu). You may also use the University's bias incident reporting system, which you can reach at https://osas.wisc.edu/report-an-issue/bias-or-hate-reporting/.

• Grade Appeals

  The Department of Statistics has a process whereby students may appeal their final letter grades in Statistics courses on grounds that the grade was calculated incorrectly or inconsistently based on the grading standards stated in this syllabus. This process is in alignment with the College of Letters and Sciences policy upholding the right of students to appeal their final course grades.

  If you believe you received an incorrect final letter grade in this course because the syllabus grading policies were applied incorrectly or inconsistently, then the first step of the process is to appeal your final grade to me. Your appeal must be:

  1. In writing
  2. Submitted within 3 weeks of the Office of the Registrar posting the grades
  3. Must explain in detail why you think the syllabus grading standards were applied incorrectly or inconsistently
  4. Include an explanation of the grade you believe you should have received had these standards been applied correctly and/or consistently

  I will respond in writing to all timely appeals made on valid grounds.

  Should I deny your appeal, you have the option to take your appeal to the Statistics Department's Appeals Committee.

  Note: this appeals process applies to final letter grades only. It is not intended for requesting regrades on individual assignments or exams! To do so, you may contact me or the TA directly. For other kinds of concerns, please check out the Complaints section of this syllabus, or contact me directly.

• Overlapping course time statement

  The Department of Statistics strongly discourages students from enrolling in any courses whose regular class meeting dates & times overlap with each other. This policy is in alignment with the College of Letters and Sciences Course Attendance Policy. It is also consistent with the Class Attendance Policy for Students at UW-Madison (https://kb.wisc.edu/ls/24628), whose first sentence reads, "It is expected that every student will be present at all classes." Statistics instructors may opt not to make any alternative arrangements in the event any conflict arises due to a student taking a course with class meetings that overlap with a Statistics course, including a conflict between two Statistics courses. Note that final exams occasionally are scheduled simultaneously for courses which meet at different times; in this situation, please contact your instructor well before the exam date about potential accommodations.

• Statement on the use of ChatGPT and other generative artificial intelligence (AI) models

  While the Statistics Department recognizes the potential benefits of AI, its use in academic work can be problematic. In this course, two rules regarding the use of ChatGPT and other generative AI models will be enforced: (1) Passing off AI-generated responses as original student work constitutes plagiarism and is strictly prohibited. Any students found to be engaging in this practice will be cited for academic misconduct. (2) Unless explicitly authorized by the instructor to do so, any use of AI-generated responses as sources of information, even with documentation and attribution, is prohibited.