No free lunch in search and optimization
No hay almuerzo gratis en búsqueda y optimización
- Date
- 1997 · confidence low
- Place
- Santa Fe, New Mexico, United States
- Credited to
- David H. Wolpert and William G. Macready
- Field
- Knowledge & symbols
What was discovered
A theorem showing that, averaged over all possible problems in a class, no search or optimization algorithm outperforms another in general: there are no universal «shortcuts», only specializations for specific problem classes.
What it led to
- Theoretical justification for domain-specialized algorithms
- Boost to the design of metaheuristics tailored to specific problems
- Reference framework for evaluating machine learning and optimization algorithms