2022/02/09 by Tom F. Sterkenburg, Peter D. Grünwald · 1 citation
Computer Science · #cs.LG
paper · pdf · doi:10.1007/s11229-021-03233-1
published as Synthese 199:9979-10015 (2021)
arxiv created 2022/02/09 · arxiv updated 2022/02/10
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.