Course Description

In this course students will gain exposure to practical aspects of machine learning and data analysis. Through a mix of lectures, student presentations, and assignments, the course will cover the various stages in modern data analysis pipelines, as well other relevant applied learning topics, including properly evaluating statistical methods, and modern societal problems related to FAT (fairness, accountability, transparency).

Course Goals

• Appreciate the challenges in practical data analysis beyond statistical modeling / prediction
• Understand proper methods for empirical evaluation of new statistical methods
• Be exposed to modern societal problems related to FAT (fairness, accountability, transparency)
• Improve communication / presentation skills

Lectures

Mondays and Wednesdays 4:30PM - 5:50PM @ GHC4303

Course Staff

• Ameet Talwalkar (Instructor)
  • Email: talwalkar@cmu.edu
  • Office Hour: By appointment
• Sebastian Caldas (TA)
  • Email: scaldas@cmu.edu
  • Office Hour: Thursday 1:00pm - 2:00pm
• Nicholay Topin (TA)
  • Email: ntopin@andrew.cmu.edu
  • Office Hour: Wednesday 3:00pm - 4:00pm
• Roy Li (TA)
  • Email: tianqil1@andrew.cmu.edu

Grading

• Assignments: 50%
• Presentations: 25%
• Presentation Write-ups: 12.5%
• Attendence / Participation: 12.5%

Attendance

Attendance to all lectures and to your own subgroup meetings are mandatory. For example, if you are in SG1, you must attend all SG1 meetings whether you are presenting or not.

Assignment Policy

Physical copies of assignments must be submitted to Christy Melucci's office (GHC 8004) between 2pm and 5pm the day they are due to avoid using late days or receiving penalties. If for some reason, you are unable to physically submit your assignment between 2pm and 5pm (due to class conflict or other issues) please email Roy (email address listed above) as soon as possible to arrange other hand-in options.

Every student gets 3 late days total (without penalty) for the duration of the class. However, no more than 2 late days can be used for a given assignment. Any late submissions beyond 2 days or exceeds the total number of late days will not receive any credit. If you are using late days, you must email the assignment to all the TAs and the instructor, with an explicit note about late days used. Late days will be 24 hour periods starting from the deadline. So, if the deadline is 5PM Monday, anything till 5PM Tuesday is 1 late day.

Subgroup Write-ups

Prior to the start of the subgroup meetings, each member of the subgroup should read each article/paper to be presented, and write a short summary of each. The write-ups are due immediately before the meetings, and you only need to do the write-ups for the articles/papers to be presented that day. You do not need to do a write-up for your own presentation. So, before the start of each subgroup meeting, the 5 presenters in that meeting must do 4 write-ups each and the 5 other members must do 5 write-ups each.

For FAT, answer each of the following questions in 2-3 sentences:

• What was the main topic/story/issue discussed in the article?
• What is the broader societal significance of this topic/story/issue?
• What is your viewpoint on this topic/story/issue?

For Experimental Evaluation, answer each of the following questions in 2-3 sentences:

• What is the general ML problem being studied in this paper?
• What is the purpose / stated goal(s) of the empirical evaluation?
• What are a few positive and/or negative aspects of the experimental evaluation?

Resources

Piazza