Past Event! Note: this event has already taken place.
RADS Seminar: From Linear Programming to Machine Learning
November 30, 2020 at 10:00 AM to 11:00 AM
Location: | Online via Zoom |
Audience: | Anyone |
Key Contact: | RADS Director |
Contact Email: | majumdar@sce.carleton.ca |
John Chinneck will be giving an online seminar titled “From Linear Programming to Machine Learning.”
Seminar Abstract:
Infeasibility arises in linear programming when a subset of the constraints conflicts. One way to analyze the problem is to find the largest subset of constraints that admits a feasible solution (maxFS), or equivalently, find the smallest subset of constraints to remove such that a feasible set is left (minURL). These problems are NP-hard, but there are good heuristics for then. Surprisingly, the problem of finding the most accurate binary linear classifier can be transformed into a maxFS/minURL problem, as can the problems of feature selection, sparse recovery in compressed sensing, nonnegative matrix factorization etc. The seminar gives an overview of the essential ideas.