Elements of information theory
Elements of information theory
The unknowable
No free lunch theorems for optimization
IEEE Transactions on Evolutionary Computation
Practical performance models of algorithms in evolutionary program induction and other domains
Artificial Intelligence
| Hi-index | 0.00 |
Three theoretical perspectives upon conservation of performance in function optimization are outlined. In terms of statistical information, performance is conserved when the distribution of functions is such that all domain subsets of a given size have identically distributed random values. In terms of algorithmic information, performance is conserved because almost all functions are algorithmically random. Also, an optimizer's algorithmic complexity bounds the information gained or lost in its processing of the test function. The practical consequences of these theoretical results are explored, with emphasis upon the results from algorithmic information theory, which are new.