CIS 677 Topics in Algorithms

Algorithmic Decision Theory and Bayesian Optimization

Class location:

Levine 612, TuTh 430-600pm.

Course Description:

We will look at optimization algorithms for problems with partial and incomplete information. These problems have a longstanding history in statistics, control and optimization theory; however are frequently computationally intractable. We will consider these problems through the lens of Approximation Algorithms, and incorporate a variety of computational constraints and not just available information or running time.

Plan:

I will be giving most of the lectures. The students will be asked to form small groups of size two or three. Each group chooses a problem in consultation with me (obviously with some connection to the course material, but not necessarily the exact same problems). The group works on the problem throughout the semester and is graded together. The grade is the smallest component of this course - the goal of the course is to learn these topics and explore possible avenues of research.

You can choose an experimental/exploratory project, but I would encourage it to have a sizeable analysis component.

Ashish Goel is teaching a similar class in Stanford and we will try to have coordinated lecture notes and supporting material.

Kamesh Munagala taught a similar class at Gatech, and the reading list there would have overlap with this course.

Papers and Topics