This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 20.3 Evaluation 321 runs, we can compute the number # q of experiments that fulfilled the predicates attached to q and the estimators of the minimum hatwide q , mean q , maximum hatwide q , median med( q ), and the standard deviation s [ q ], and so on. Obviously, not all of them are needed or carry a meaning in every experiment. Table 20.5 lists some of these first metrics. 322 20 Experimental Settings, Measures, and Evaluations Measure Short Description Number of Successful Runs # s The number of runs where successful individuals were discovered. # s = { i : ( s τ i negationslash = ∅ ) ∧ (0 ≤ i < # r ) } (20.1) Success Fraction s / r The fraction of experimental runs that turned out successful. s / r = # s # r (20.2) Minimum Success Evaluation hatwider s τ The number of evaluations τ needed by the fastest (successful) ex perimental run to find a successful individual. (or ∅ if no run was successful) hatwide s τ = braceleftbigg min { s τ i negationslash = ∅} if ∃ s τ i negationslash = ∅ ∅ otherwise (20.3) Mean Success Evaluation s τ The average number of evaluations τ needed by the (successful) experimental runs to find a successful individual. (or ∅ if no run was successful) s τ = braceleftBigg ∑ s τ i negationslash = ∅ s τ i { s τ i negationslash = ∅} if ∃ s τ i negationslash = ∅ ∅ otherwise (20.4) Maximum Success Evaluation hatwider s τ The number of evaluations τ needed by the slowest (successful) experimental run to find a successful individual. (or ∅ if no run was successful) hatwide s τ = braceleftbigg max { s τ i negationslash = ∅} if ∃ s τ i negationslash = ∅ ∅ otherwise (20.5) Minimum Success Generation hatwide s t The number of generations/iterations t needed by the fastest (suc cessful) experimental run to find a successful individual. (or ∅ if no run was successful) hatwide s t = braceleftbigg min { s t i negationslash = ∅} if ∃ s t i negationslash = ∅ ∅ otherwise (20.6) Mean Success Generation s t The average number of generations/iterations t needed by the (suc cessful) experimental runs to find a successful individual. (or ∅ if no run was successful) s t = braceleftBigg ∑ s t i negationslash = ∅ s t i { s t i negationslash = ∅} if ∃ s t i negationslash = ∅ ∅ otherwise (20.7) Maximum Success Generation hatwide s t The number of Generations generations/iterations t needed by the slowest (successful) experimental run to find a successful individual. (or ∅ if no run was successful) hatwide s t = braceleftbigg max { s t i negationslash = ∅} if ∃ s t i negationslash = ∅ ∅ otherwise (20.8) Number of Perfect Runs # p The number of runs where perfect individuals were discovered....
View
Full
Document
This document was uploaded on 08/10/2011.
 Spring '11
 Algorithms

Click to edit the document details