To Explain or to Predict? — mentioned by ModelTheorist12

Wed, 16 Oct 2019 14:50:25 GMT

Given a finite data set, an infinite number of functions/curves can explain it. This is curve-fitting 101 stuff. All explanatory models are either an over-fit or an under-fit to the data. With the fine print being "acceptable error rate". "The Earth is flat" may be wrong, but it could be acceptably wrong to me. This is tackled by model-dependent realism ( https://en.wikipedia.org/wiki/Model-dependent_realism ) and by this paper "To explain or to predict" ( https://arxiv.org/pdf/1101.0891.pdf )

Paper and discussion history. A mention is not an endorsement.

To Explain or to Predict? · HackerPapers discussions

To Explain or to Predict? — mentioned by ModelTheorist12