Forster and Sober on the Curve-Fitting Problem
暂无分享,去创建一个
Forster and Sober present a solution to the curve-fitting problem based on Akaike's Theorem. Their analysis shows that the curve with the best epistemic credentials need not always be the curve that most closely fits the data. However, their solution does not, without further argument, avoid the two difficulties that are traditionally associated with the curve-fitting problem: (1) that there are infinitely many equally good candidate-curves relative to any given set of data, and (2) that these best candidates include curves with indefinitely many bumps.
[1] Malcolm R. Forster,et al. How to Tell When Simpler, More Unified, or Less Ad Hoc Theories will Provide More Accurate Predictions , 1994, The British Journal for the Philosophy of Science.
[2] H. Akaike. A new look at the statistical model identification , 1974 .
[3] H. Akaike,et al. Information Theory and an Extension of the Maximum Likelihood Principle , 1973 .