Leo Breiman

Leo Breiman (1928–2005) was an American statistician who transformed machine learning from a branch of statistics into a field in its own right. He is best known for inventing Bagging (Bootstrap Aggregating) in 1996 and Random Forest in 2001, two methods that turned the instability of decision trees into a source of strength through ensemble design. Before his work on machine learning, Breiman made fundamental contributions to probability theory, information theory, and applied statistics.

Breiman was a vocal critic of the statistical establishment's obsession with model interpretability at the expense of predictive accuracy. His 2001 paper "Statistical Modeling: The Two Cultures" argued that the data modeling culture — which assumes a true model exists and searches for it — had been superseded by the algorithmic modeling culture, which treats models as black-box prediction engines. This was not a technical claim but a disciplinary one: Breiman was declaring that the future of statistics belonged to machines, not to parameters.