Ben Recht 12/9/2025

There is no data-generating distribution

Read Original

The article is a reflection from a professor's final lecture on machine learning, arguing against the standard assumption of a 'data-generating distribution.' It discusses how removing this and other field-making myths from the curriculum leads to a more honest, distribution-free understanding of ML, focusing on populations, samples, and the engineer's role in creating or imagining randomness.

There is no data-generating distribution

Comments

No comments yet

Be the first to share your thoughts!

Browser Extension

Get instant access to AllDevBlogs from your browser