David Hogg, NYU
When
Abstract: Machine learning (ML) methods are having a huge impact across all of the sciences. However, ML has a strong ontology—in which only the data exist—and a strong epistemology—in which a model is considered good if it performs well on held-out training data. These philosophies are in strong conflict with both standard practices and key philosophies in astrophysics. I show that there are contexts in which the introduction of ML introduces strong, unwanted (and currently unfixable) statistical biases. However, there are locations for ML in astrophysics in which the ontology and epistemology are valuable: I will even give an example of a place where the introduction of ML makes your project or measurement or conclusion more conservative, reliable, and trustworthy (no cap).