翻訳と辞書 Words near each other ・ Evolutionary systems ・ Evolutionary taxonomy ・ Evolutionary Technologies International ・ Evolutionary theory and the political left ・ Evolutionary trade-offs ・ Evolutionary trap ・ Evolutionism ・ Evolutionist Liberal Party ・ Evolutionist Liberal Party of Ceará ・ Evolutionist Party ・ Evolutions Television ・ Evoluzione della specie ・ Evolv ・ Evolva ・ Evolvability ・ Evolvability (computer science) ・ Evolvable hardware ・ Evolve ・ Evolve (Ani DiFranco album) ・ Evolve (coldrain album) ・ Evolve (Endo album) ・ Evolve (EP) ・ Evolve (Eye Empire album) ・ Evolve (professional wrestling) ・ Evolve (song) ・ Evolve (TV series) ・ Evolve (video game) ・ Evolve 4.0 ・ Evolve Championship ・ Evolve Festival Dictionary Lists mini英和辞書 mini和英辞書 Webster 1913 Latin-English FOLDOC Wikipedia English ウィキペディア

翻訳と辞書　辞書検索 [ 開発暫定版 ] スポンサード リンク Evolvability (computer science) ： ウィキペディア英語版 Evolvability (computer science) The term evolvability is used for a recent framework of computational learning introduced by Leslie Valiant in his paper of the same name and described below. The aim of this theory is to model biological evolution and categorize which types of mechanisms are evolvable. Evolution is an extension of PAC learning and learning from statistical queries. ==General Framework== Let $F\_n\backslash ,$ and $R\_n\backslash ,$ be collections of functions on $n\backslash ,$ variables. Given an ''ideal function'' $f\; \backslash in\; F\_n$, the goal is to find by local search a ''representation'' $r\; \backslash in\; R\_n$ that closely approximates $f\backslash ,$. This closeness is measured by the ''performance'' $\backslash operatorname(f,r)$ of $r\backslash ,$ with respect to $f\backslash ,$. As is the case in the biological world, there is a difference between genotype and phenotype. In general, there can be multiple representations (genotypes) that correspond to the same function (phenotype). That is, for some $r,r\text{'}\; \backslash in\; R\_n$, with $r\; \backslash neq\; r\text{'}\backslash ,$, still $r(x)\; =\; r\text{'}(x)\backslash ,$ for all $x\; \backslash in\; X\_n$. However, this need not be the case. The goal then, is to find a representation that closely matches the phenotype of the ideal function, and the spirit of the local search is to allow only small changes in the genotype. Let the ''neighborhood'' $N(r)\backslash ,$ of a representation $r\backslash ,$ be the set of possible mutations of $r\backslash ,$. For simplicity, consider Boolean functions on $X\_n\; =\; \backslash ^n\backslash ,$, and let $D\_n\backslash ,$ be a probability distribution on $X\_n\backslash ,$. Define the performance in terms of this. Specifically, :$\backslash operatorname(f,r)\; =\; \backslash sum\_\; f(x)\; r(x)\; D\_n(x).$ Note that $\backslash operatorname(f,r)\; =\; \backslash operatorname(f(x)=r(x))\; -\; \backslash operatorname(f(x)\; \backslash neq\; r(x)).$ In general, for non-Boolean functions, the performance will not correspond directly to the probability that the functions agree, although it will have some relationship. Throughout an organism's life, it will only experience a limited number of environments, so its performance cannot be determined exactly. The ''empirical performance'' is defined by $\backslash operatorname\_s(f,r)\; =\; \backslash frac\; \backslash sum\_\; f(x)r(x),$ where $S\backslash ,$ is a multiset of $s\backslash ,$ independent selections from $X\_n\backslash ,$ according to $D\_n\backslash ,$. If $s\backslash ,$ is large enough, evidently $\backslash operatorname\_s(f,r)$ will be close to the actual performance $\backslash operatorname(f,r)$. Given an ideal function $f\; \backslash in\; F\_n$, initial representation $r\; \backslash in\; R\_n$, ''sample size'' $s\backslash ,$, and ''tolerance'' $t\backslash ,$, the ''mutator'' $\backslash operatorname(f,r,s,t)$ is a random variable defined as follows. Each $r\text{'}\; \backslash in\; N(r)$ is classified as beneficial, neutral, or deleterious, depending on its empirical performance. Specifically, * $r\text{'}\backslash ,$ is a beneficial mutation if $\backslash operatorname\_s(f,r\text{'})\; -\; \backslash operatorname\_s(f,r)\; \backslash geq\; t$; * $r\text{'}\backslash ,$ is a neutral mutation if ; * $r\text{'}\backslash ,$ is a deleterious mutation if $\backslash operatorname\_s(f,r\text{'})\; -\; \backslash operatorname\_s(f,r)\; \backslash leq\; -t$. If there are any beneficial mutations, then $\backslash operatorname(f,r,s,t)$ is equal to one of these at random. If there are no beneficial mutations, then $\backslash operatorname(f,r,s,t)$ is equal to a random neutral mutation. In light of the similarity to biology, $r\backslash ,$ itself is required to be available as a mutation, so there will always be at least one neutral mutation. The intention of this definition is that at each stage of evolution, all possible mutations of the current genome are tested in the environment. Out of the ones who thrive, or at least survive, one is chosen to be the candidate for the next stage. Given $r\_0\; \backslash in\; R\_n$, we define the sequence $r\_0,r\_1,r\_2,\backslash ldots$ by $r\_\; =\; \backslash operatorname(f,r\_i,s,t)$. Thus $r\_g\backslash ,$ is a random variable representing what $r\_0\backslash ,$ has evolved to after $g\backslash ,$ ''generations''. Let $F\backslash ,$ be a class of functions, $R\backslash ,$ be a class of representations, and $D\backslash ,$ a class of distributions on $X\backslash ,$. We say that $F\backslash ,$ is ''evolvable by $R\backslash ,$ over $D\backslash ,$'' if there exists polynomials $p(\backslash cdot,\backslash cdot)$, $s(\backslash cdot,\backslash cdot)$, $t(\backslash cdot,\backslash cdot)$, and $g(\backslash cdot,\backslash cdot)$ such that for all $n\backslash ,$ and all $\backslash epsilon\; >\; 0\backslash ,$, for all ideal functions $f\; \backslash in\; F\_n$ and representations $r\_0\; \backslash in\; R\_n$, with probability at least $1\; -\; \backslash epsilon\backslash ,$, :$\backslash operatorname(f,r\_)\; \backslash geq\; 1-\backslash epsilon,$ where the sizes of neighborhoods $N(r)\backslash ,$ for $r\; \backslash in\; R\_n\backslash ,$ are at most $p(n,1/\backslash epsilon)\backslash ,$, the sample size is $s(n,1/\backslash epsilon)\backslash ,$, the tolerance is $t(1/n,\backslash epsilon)\backslash ,$, and the generation size is $g(n,1/\backslash epsilon)\backslash ,$. $F\backslash ,$ is ''evolvable over $D\backslash ,$'' if it is evolvable by some $R\backslash ,$ over $D\backslash ,$. $F\backslash ,$ is ''evolvable'' if it is evolvable over all distributions $D\backslash ,$. 抄文引用元・出典: フリー百科事典『 ウィキペディア（Wikipedia）』 ■ウィキペディアで「Evolvability (computer science)」の詳細全文を読む スポンサード リンク 翻訳と辞書 : 翻訳のためのインターネットリソース Copyright(C) kotoba.ne.jp 1997-2016. All Rights Reserved.