Extending Instance Space Analysis to Algorithm Configuration Spaces

What is it about?

This paper is about developing methodologies for visualising how parameter configurations (of algorithms) perform in terms of the characteristics (features) of problem instances (to which they are applied). Parameter configurations and problem instances are projected to two related spaces, and the relationship between the two spaces can be used to obtain insights.

Featured Image

Photo by Ben Wicks on Unsplash

Photo by Ben Wicks on Unsplash

Why is it important?

This paper represents initial work in extending a methodology for algorithmic insights, known as instance space analysis, to incorporate algorithm parameters directly into the analysis.