Dynamic Programming Approach to Decision Rule Optimization: Problems of Discretization

Dynamic Programming Approach to Decision Rule Optimization: Problems of Discretization

Internship Description

The dynamic programming approach to decision rule optimization created in KAUST allows optimizing of decision rules relative to length and coverage, and study of relationships between two cost functions or between cost and accuracy of decision rules. This approach is applicable to data sets with discrete attributes only. To apply it to datasets with continuous attributes it is necessary to make of discretization of these attributes either before or during the process of optimization. The aim of internship is to compare different approaches to discretization from the point of view of usefulness for knowledge representation and machine learning.

Faculty Name

Field of Study

Combinatorial Optimization, Machine Learning, Knowledge Representation