In machine learning, a common problem is classification of individuals into groups based on their features. A model might be trained to recognize patterns in the features of a member of a population; however, more does not always equal better. The observation of increasingly many different features eventually introduces confounding variables and statistical noise into the model, so it is often better to verify if a feature is hurting the accuracy of the model and eliminate it from consideration accordingly.

Feature Selection begins with a frantic oscillation between multiple different textures. As the piece runs its course, the textures are gradually pruned. By the end, only a small fragment is left of the array of materials presented in the opening.

​

Written for the first "48 Hours" program at IU.

Selected for SCI 2021 National Conference Call for Works