schofield2020evolving.pdf (358.98 kB)

Evolving Simpler Constructed Features for Clustering Problems with Genetic Programming

Download (358.98 kB)
conference contribution
posted on 12.08.2021, 03:17 by F Schofield, Andrew LensenAndrew Lensen
Clustering is a widely used unsupervised learning technique. However, as the size and complexity of data increases, the performance of clustering algorithms diminishes, as well as the interpretability of the clustering partition. Genetic programming has been used to perform feature construction on data to increase clustering performance. However, existing work has not focused on encouraging simpler constructed features. In this paper, existing techniques are further developed to include parsimony pressure-a method to encourage evolution towards simpler solutions. With simpler solutions, the constructed features become easier to understand and interpret. The results of experiments using the proposed method show that parsimony pressure is an effective method for producing significantly simpler constructed features without any reduction on the performance of k-means++clustering. Evolved individuals are also analysed to demonstrate the effect of parsimony pressure on interpretability, showing the power of parsimony pressure for avoiding redundancies in individuals, and thus increasing the interpretability.

History

Preferred citation

Schofield, F. & Lensen, A. (2020, July). Evolving Simpler Constructed Features for Clustering Problems with Genetic Programming. In 2020 IEEE Congress on Evolutionary Computation, CEC 2020 - Conference Proceedings 2020 IEEE Congress on Evolutionary Computation (CEC) (00 pp. 1-8). IEEE. https://doi.org/10.1109/CEC48606.2020.9185575

Conference name

2020 IEEE Congress on Evolutionary Computation (CEC)

Conference start date

19/07/2020

Conference finish date

24/07/2020

Title of proceedings

2020 IEEE Congress on Evolutionary Computation, CEC 2020 - Conference Proceedings

Volume

00

Publication or Presentation Year

01/07/2020

Pagination

1-8

Publisher

IEEE

Publication status

Published

Usage metrics

Categories

Exports