This project is read-only.
Project Description
The Graph Properties Extractor is a friendly GUI that allows the user to extract numeric properties from graphs.

Project Background
The Algorithm Selection Problem introducted by Rice (1975) is defined as the following:

P: the Problem Space representing a set of instances of a Problem,
A: the Algorithm Space containing all the algorithms for tackling this Problem, and
Y: the Performance Space containing the performance metrics for the Algorithms in A,
Find the Selection Mapping S(x) function that selects an Algorithm from A maximizing the performance Y.

Metalearning applied on the problem of Algorithm Selection is currently viewed as the application of Machine Learning algorithms to generate metaknowledge mapping characteristics of problems (metafeatures) to the relative performance of algorithms.

Project Scope
This Project will consider the Graph Coloring Problem as a case study of Metalearning applied to the Algorithm Selection Problem.

The system will receive a Graph Coloring Problem set of instances as input to generate a model using Machine Learning approach to learn about the algorithms performances.

With the learned model, the system can then predict for an unseen instance, what is the algorithm with the expected best performance.

Related Works
Other researchers have already developed work considering the Graph Coloring Problem as a case study, but no common methodology has been employed.

For more information go to our Documentation page.

Last edited Jul 22, 2014 at 12:59 AM by lucasdeandrade87, version 8