This doctoral-level seminar will introduce the
statistical and algorithmic principles in the field of scalable machine
learning. The course will consist of instructor-led lectures and
student-led presentations on current research topics.
This course is intended for PhD students with interest
in working on research problems in areas related to scalable machine learning.
Students are expected to have a strong background in
machine learning at the graduate level (CS260 or equivalent), and
also have a solid background in statistics, optimization, and linear
algebra. Permission from the instructor is required to enroll. If you think you are qualified
for this course, please email Prof. Talwalkar with detailed
information about your past training in machine learning (both coursework and research experience).
Grades will be based on the following components:
- Presentation (35%)
- Each student is required to prepare and present a
lecture on a current research topic in the area of scalable machine
learning. This will involve reading several research papers
and distilling the core ideas into a 1.5 hour technical presentation.
- The presenting student must select a short reading assignment
for the class, and send it to the entire class at least one week
before the scheduled presentation.
- Please carefully read these tips for preparing your class presentation.
- Participation (15%)
- Students are expected to attend all classes, complete all course
readings prior to class, and actively participate in class discussions.
- Project (50%)
- Students can work alone or in small groups.
- Deliverables include:
- Initial Proposal due April 17: 1 page writeup along with a short in-class
- Progress Report due May 10: 1 page
- Final Report due June 7: 5 page
writeup along with an in-class
- It is expected that several class projects will form the basis
of future publications. Projects will thus be evaluated as if they
are submissions to a technical workshop/conference.
Topics and Tentative Schedule