Thomas Lumley • 12/30/2016

The ‘iris’ data

The article critiques the famous 'iris' dataset, arguing it's a poor example for modern machine learning due to its small size and low dimensionality. It reveals the dataset's original purpose was not for classification/clustering practice, but for a specific biological hypothesis testing whether Iris versicolor was a hybrid, as discussed by Fisher and Anderson.

0 comments

#Machine Learning #data visualization #classification