Course Syllabus

Instructors:
 David Koes 
 748 Murdoch Building
dkoes@pitt.edu
 
 Maria Chikina
 833 Murdoch Building
 mchikina@pitt.edu

Teaching Assistant:

Thomas Jacob
thj19@pitt.edu

Course Description

This course will focus on the practical aspects of effectively applying state-of-the-art machine learning methods to biomedically relevant datasets. Topics covered include mathematical foundations, practical coding skills, classical machine learning, deep learning, and generative modeling.

 
High-throughput techniques are revolutionizing biomedical research. From whole genome sequencing, to RNA-Seq transcriptome profiling, to high-throughput mass spectrometry for protein profiling, to high-throughput biochemical screening, to flow cytometry for cell profiling, to high-content screening, to literature analysis and electronic medical records, from molecule to patient, modern techniques generate vast quantities of data. In order to be effective, biomedical researchers require the appropriate computational tools to correctly interpret and utilize this data. As machine learning is the science of finding and applying patterns in data, it is an essential tool for turning data into knowledge and actionable insights and has been rising in prominence in biomedical research.

Communication

Course material will be posted to Canvas: https://canvas.pitt.edu/courses/310776. Slack  will be used for group discussions, announcements, and contacting staff: http://cobb2060.slack.com. 

Lectures 

Lectures will be noon-1:20pm on Tuesdays and Thursdays.  Lectures will be in-person in the Murdoch 814 classroom.  If you do not already have access to the Murdoch building, fill out this form.

Recitations

Recitations will be 12:30pm - 1:20pm on Wednesdays in the Murdoch 814 classroom.  Recitations will be a mix of practical, in-class projects and lectures.  They are not optional.

Class Recordings

All lectures and recitations will be recorded and available for asynchronous viewing on Panopto, but this is not intended as a substitute for attending class. Students should make every reasonable effort to attend class in real-time as in-class group work and discussion is an important part of lecture.  

Assignments

There will be 8 assignments. They are the heart and soul of the course.  Most assignments will take the form of a "friendly competition" where students submit their solutions and have their (anonymous) results posted to a leaderboard.  There will be preset thresholds for achieving an "A" (95) quality solution.  Students work individually on the assignments.  Undergraduates may choose to work on assignments in groups with the permission of the instructors.

Quizzes

There will be two in-class quizzes.

Grades

A final overall score of at least 93% will be required for an A and at least 83% for a B.  Assignment grades will be based on the final status of the leaderboard (not the best scoring solution submitted).

Lateness 

Assignments should be handed in on-time.  When this is not possible, course instructors should be contacted with as much advance notice as possible.  In general, requests for one late day per an assignment for no more than two total assignments will be approved.  Requests beyond that will require substantial justification and/or be subject to additional grade penalties.  Late assignments will have a maximum possible score of 95%.

Academic Honesty 

You must do all your own work.  You are encouraged to discuss general concepts, strategies for debugging, and the particulars of a specific software package with other class members.  However, specifics of individual assignments should not be discussed, and you should not show your code to fellow classmates.  You are expected to understand and be able to explain any code you submit. Any attempt to "hack" the autograder will result in expulsion from the class and a referral to the dean's office.