The no-free-lunch theorems of supervised learning
This addresses a foundational philosophical issue in machine learning for theorists, but it is incremental as it builds on existing debates.
The paper tackles the problem of reconciling no-free-lunch theorems with learning theory by arguing that algorithms are model-dependent, not purely data-driven, allowing for model-relative justification.
The no-free-lunch theorems promote a skeptical conclusion that all possible machine learning algorithms equally lack justification. But how could this leave room for a learning theory, that shows that some algorithms are better than others? Drawing parallels to the philosophy of induction, we point out that the no-free-lunch results presuppose a conception of learning algorithms as purely data-driven. On this conception, every algorithm must have an inherent inductive bias, that wants justification. We argue that many standard learning algorithms should rather be understood as model-dependent: in each application they also require for input a model, representing a bias. Generic algorithms themselves, they can be given a model-relative justification.