UROP Project

Attacking machine learning algorithms with optimization

Contact

Name

Daniel Holder

Program Director UROP

Telephone

workPhone
+49 241 80-90695

E-Mail

Key Info

Basic Information

Project Offer-Number:
833
Category:
RWTH UROP, UROP Network, UROP International
Field:
Mathematics
Faculty:
8
Organisation unit:
Operations Research
Language Skills:
English or German
Computer Skills:
strong data and implementation skills required

MoveOn

In the era of ubiquitous availability of data, machine learning gained renewed interest in order to "make sense" of them. Many machine learning tasks bear in fact an optimization component; one is looking for a best classification, a best clustering, or a best explanation, etc. We would like to extract and formulate these tasks as optimization models (like integer programs) and try to solve them as such.

Task

A concrete machine learning approach (like using a support vector machine, looking a clustering algorithms, etc.) will be considered, the optimization problem extracted, a mathematical optimization model formulated, and solved. This may require exploiting a (to be identified) mathematical structure of the machine learning task.

Requirements

A strong general computer affinity (scripting languages, data bases, etc.) is very welcome; knowledge in machine learning and/or mathematical optimization should be at least basic