Skip over navigation

Course Offerings

Course Evaluation Results

Course Details

Spring 2014-2015
COS 598D   Graded A-F, P/D/F, Audit

Advanced Topics in Computer Science - Overcoming computational intractability in machine learning

Sanjeev Arora

This seminar course focuses on myriad machine learning problems that are computationally intractable (NP-hard or worse). Often the standard approaches used in practice are heuristics with no provable bounds on running times or solution quality. Is this necessary? How can we develop an algorithmic theory in these settings with provable run times? This requires good modeling as well as algorithm design. We'll cover several recent successes along these lines, and also think about open problems.

Sample reading list:
See instructor for complete list

Reading/Writing assignments:
Auditors welcome but should contact the instructor.

Other Requirements:
Not Open to Freshmen.

Prerequisites and Restrictions:
Prerequisites: Upper level course in theoretical CS, preferable algorithms. Undergrad level knowledge of machine learning/AI. Otherwise instructor's permission..

Other information:
Class format: Lectures, discussions.

Schedule/Classroom assignment:

Class numberSectionTimeDaysRoomEnrollmentStatus
43310 S01 1:30 pm - 4:20 pm Th   Computer Science Building   302   Enrolled:11 Limit:20