Fiche du document
2017
- handle: 10670/1.muulxo
- halshs: halshs-01474531
- doi: 10.1007/978-3-319-54469-4_13
Ce document est lié à :
info:eu-repo/semantics/altIdentifier/doi/10.1007/978-3-319-54469-4_13
info:eu-repo/semantics/OpenAccess
Mots-clés
Data Analysis Optimality Forcing Brandt's principleSujets proches
Instruction and study--Methods Instruction and study--MethodsCiter ce document
Domenico Napoletani et al., « Forcing optimality and Brandt's principle », HAL-SHS : histoire, philosophie et sociologie des sciences et des techniques, ID : 10.1007/978-3-319-54469-4_13
Métriques
Partage / Export
Résumé
We argue that many optimization methods can be viewed as representatives of " forcing " , a methodological approach that attempts to bridge the gap between data and mathematics on the basis of an a priori trust in the power of a mathematical technique, even when detailed, credible models of a phenomenon are lacking or do not justify the use of this technique. In particular, we show that forcing is implied in particle swarms optimization methods, and in modeling image processing problems through optimization. From these considerations, we extrapolate a principle for general data analysis methods, what we call 'Brandt's principle', namely the assumption that an algorithm that approaches a steady state in its output has found a solution to a problem, or needs to be replaced. We finally propose that biological systems, and other phenomena that respect general rules of morphogene-sis, are a natural setting for the application of this principle.