Trees of Discomfort

Trees of Discomfort

2022 Brussels Constant Vitrine Mixed Media with micro-controllers (ESP-32)

Partners : Gijs de Heij, Anaïs Berck, An Mertens

Support : Constant, FRART, Meise Botanical garden

Photo credits : Guillaume Slizewicz

An artistic interpretation of the random forest algorithm.

Trees of Discomfort is an artistic interpretation of the random forest algorithm, a machine learning algorithm in which several decision trees are trained on a dataset and form a forest. The trees of the of forest try to predict the category of a given observation. They do this by answering a set of yes or no questions. Starting at the root of the tree they follow either the left or right branch to arrive at the next question (a knot), or at an answer (an endpoint). Once the trees have decided for themselves they vote and the forest comes to a decision.

In Trees of discomfort five (decision-)trees and an anouncer live on six micro-controllers (ESP-32). When an observation is announced each of the trees will try to classify the observation and explains its thinking.

Once in a while, a tree expresses doubts on its behaviour and the context in which it was developed and voices its unease on botany, herbaria, the iris dataset, and processes of classification.

Algorithm: Random Forest

Datasets: Iris Dataset, Living Herbarium of Meise, Digital Herbarium of Meise.

Humans: Gijs de Heij, Guillaume Slizewicz

You can find the code of this installation at: