進(jìn)化算法的設(shè)計(jì)與應(yīng)用研究

前言

This book covers broad spectrum important subjects ranging from generalinterests of EAs （such as algor！thm parameter control and constraint handling） tothe hottest topics in EAs-estimation of distribution algorithms （EDAs） withfocusing on design and applications of EDAS.The book is comprised of total of 6 chapters. In Chapter 1, we discuss how toset the values of various parameters of an evolutionary algorithm, beginning withthe issue of whether these values are best set in advance or are best changed duringevolution. Then, we provide a classification of different approaches based on anumber of complementary features, and pay special attention to setting parameterson-the-fly. Then, we consider the issue of constraint handling using evolutionaryalgorithms in Chapter 2. Based on the classification of constrained problems, wediscuss what constraint handling means from an EA perspective, and study the most commonly applied EA techniques to treat constraints. However, Chapter 3 focuses on the parallelization of EDAs. More specifically, it gives guidelines for designing efficient parallel EDAs that employ parallel fitness evaluation and parallel model building. Furthermore, techniques of implementati6ns of new type of EDAs are studied in Chapter 4. Finally, Chapter 5 and Chapter 6 bring together some of EDAs approaches to optimization problems in the fields of medical science and resource management.

內(nèi)容概要

　　《進(jìn)化算法的設(shè)計(jì)與應(yīng)用研究》涵蓋了一系列重要的主題，范圍從進(jìn)化算法普遍涉及的問題（如算法參數(shù)控制與約束處理）到進(jìn)化算法的研究熱點(diǎn)——分布估計(jì)算法，并且重點(diǎn)突出分布估計(jì)算法的設(shè)計(jì)與應(yīng)用。全書共有6章。第1章討論如何設(shè)置進(jìn)化算法各種參數(shù)的值，以這些參數(shù)值是否最好事先設(shè)置或在進(jìn)化過程中如何改變等議題開始，給出了許多基于互補(bǔ)特征的不同方法的分類。第2章討論使用進(jìn)化算法時的約束處理?；诩s束問題的分類，討論從進(jìn)化算法角度理解約束處理的含義，并研究了最常用的約束處理的進(jìn)化技術(shù).第3章集中于設(shè)計(jì)并行分布估計(jì)算法，更詳細(xì)地給出了利用并行適應(yīng)度評價和并行建模設(shè)計(jì)有效分布估計(jì)算法的具體指導(dǎo)。此外，第4章研究設(shè)計(jì)一類新的分布估計(jì)算法。最后，第5章和第6章匯合了分布估計(jì)算法解決醫(yī)學(xué)和資源管理領(lǐng)域的優(yōu)化問題的研究?！哆M(jìn)化算法的設(shè)計(jì)與應(yīng)用研究》對進(jìn)化算法領(lǐng)域的研究人員來說非常有用，也可供計(jì)算機(jī)專業(yè)的博士、碩士研究生使用。

書籍目錄

1 在進(jìn)化算法中如何設(shè)置參數(shù)的值1.1 引言1.2 如何改變參數(shù)1.2.1 改變變異規(guī)模1.2.2 改變懲罰系數(shù)1.2.3 總結(jié)1.3 進(jìn)化算法參數(shù)控制技術(shù)分類1.3.1 改變算法的成分或參數(shù)1.3.2 改變參數(shù)值的方法1.3.3 決定改變參數(shù)值的依據(jù)1.3.4 改變的范圍1.3.5 總結(jié)1.4 改變進(jìn)化算法參數(shù)的案例1.4.1 表達(dá)式1.4.2 適應(yīng)度函數(shù)1.4.3 變異1.4.4 交叉1.4.5 選擇1.4.6 種群1.4.7 同時改變幾個參數(shù)1.5 討論2 進(jìn)化算法中的約束處理2.1 引言2.2 約束問題2.2.1 無約束的優(yōu)化問題2.2.2 約束滿足問題2.2.3 受約束的優(yōu)化問題2.3 約束處理的種類2.4 約束處理的途徑2.4.1 懲罰函數(shù)2.4.2 糾正函數(shù)2.4.3 限制搜尋在可行域內(nèi)2.4.4 解碼器函數(shù)2.5 應(yīng)用實(shí)例2.5.1 間接解決方法2.5.2 直接解決方法3 設(shè)計(jì)并行分布估計(jì)算法指導(dǎo)3.1 引言3.2 并行分布估計(jì)算法的方法3.2.1 分布式適應(yīng)度評價3.2.2 構(gòu)建分布式模型3.3 混合貝葉斯優(yōu)化算法3.4 復(fù)雜性分析3.4.1 選擇算子的復(fù)雜性3.4.2 構(gòu)造模型的復(fù)雜性3.4.3 模型取樣的復(fù)雜性3.4.4 替換算子的復(fù)雜性3.4.5 適應(yīng)度評價的復(fù)雜性3.5 可擴(kuò)展性分析3.5.1 處理器數(shù)為固定時的可擴(kuò)展性3.5.2 處理器數(shù)增加時可擴(kuò)展性如何變化4 基于最大熵原理設(shè)計(jì)一類新的分布估計(jì)算法4.1 引言4.2 熵、模式4.2.1 熵4.2.2 在子集條件約束下的最大熵4.2.3 模式4.2.4 最大熵分布和模式約束4.3 算法的基本思路4.4 分布估計(jì)和取樣4.5 新算法4.5.1 一階模式算法4.5.2 二階模式算法4.6 實(shí)驗(yàn)結(jié)果4.7 結(jié)論5 基于種群遞增學(xué)習(xí)算法的癌癥化療優(yōu)化技術(shù)5.1 引言5.2 癌癥化學(xué)療法的優(yōu)化問題5.2.1 化學(xué)療法的醫(yī)學(xué)處理5.2.2 癌癥化療模型5.3 GA和PBIL解決方案5.3.1 問題的編碼5.3.2 遺傳算法5.3.3 基于種群遞增學(xué)習(xí)算法5.4 實(shí)驗(yàn)結(jié)果5.4.1 算法有效性比較5.4.2 化療治療效果比較5.5 結(jié)論6 應(yīng)用分布估計(jì)算法和遺傳算法優(yōu)化動態(tài)價格問題6.1 引言6.2 通過動態(tài)價格途徑提高資源管理6.3 動態(tài)價格模型6.4 動態(tài)價格的進(jìn)化算法解決方案6.4.1 進(jìn)化算法解的表達(dá)式6.4.2 進(jìn)化算法6.5 實(shí)驗(yàn)及結(jié)果6.5.1 算法參數(shù)化6.5.2 結(jié)果6.5.3 結(jié)果分析6.6 結(jié)論

章節(jié)摘錄

插圖：optimization objectives.After the transformation,they effectively practicallydisappear, and all we need to care about is optimizing the resulting objectivefunction. This type of constraint handling is done before the EA run.（2）As an alternative to this option we distinguish direct constraint handling,meaning that the problem offered to the EA to solve has constraints （is a COP） thatare enforced explicitly during the EA run.It should be clear from the previous discussion that these options are notexclusive: for a given constrained problem （CSP or COP） some constraints mightbe treated directly and some others indirectly.It is also important to note that even when all constraints are treatedindirectly, so that we apply an EA for an FOP, this does not mean that the EA isnecessarily ignoring the constraints. In theory one could fully rely on the generaloptimization power of EAs and try to solve the given FOP without taking note ofhow f is obtained. However, it is also possible that one does take the specificorigin of f into account, i.e. the fact that is constructed from constraints. In thiscase one can try to make use of specific constraint-based information within the EAby, for instance, special mutation or crossover operators that explicitly aim atsatisfying constraints by using some heuristics.Finally, let us reiterate that indirect constraint handling is always part of thepreparation of the problem before offering it to an EA to solve. However, directconstraint handling is an issue within the EA constituting methods that enforcesatisfaction of the constraints.

編輯推薦

《進(jìn)化算法的設(shè)計(jì)與應(yīng)用研究》由華中科技大學(xué)出版社出版。

圖書封面

圖書標(biāo)簽Tags

無

評論、評分、閱讀與下載

---

    進(jìn)化算法的設(shè)計(jì)與應(yīng)用研究 PDF格式下載

---