igA

IGA 簡介

交談式(或互動式)遺傳演算法(Interactive Genetic Algorithms; IGA)主要概念是以「人」扮演GA (Genetic Algorithms)適應函數的角色。IGA系統可以描述為:一個「沒有目標函數的GA」,一個(些) 「扮演目標函數的使用者」,以及一個「使用者與GA互動的介面」組成的系統,目的是幫助使用者找出心中想要的東西。儘管與GA只有些微差異,但IGA 卻發展出一種結合「GA 全域搜尋能力」及「人類的評估力」兩者特色的適應性系統。此類型研究通稱交談式(或互動式)演化計算(Interactive Evolutionary Computation)。

本著作發表之前,IGA模型忽略了一個重要的觀念:由於IGA假設適應函數未知,因此研究人員設計IGA解答空間時,通常容易疏忽目標空間。理論上,決定解答空間的因素應該是目標空間;愈能掌握目標空間,則目標函數所對應出的解答空間將愈完整。研究IGA的學者在決定解答空間時,即使有預先考量到決策者的價值或目標,但基於下列部分或全部之理由,他們可能都只考量決策者目標集合的部分集合:

1)只關心定性目標,疏忽定量目標,

2)只考慮一般決策者的目標,忽略特殊決策者的目標,

3)只考慮多數決策者的共同目標,忽略少數決策者的目標,

4)直接考慮解答空間,完全不從目標空間思考。

目標空間是創造可行方案的基礎,如果一個完整的目標空間提供給決策者的創造力(Creativity)為1,不完整的目標空間所能提供給決策者的創造力將小於1。研究IGA的學者根據部分目標集合所決定出的解答空間,理論上是一個不完整的解答空間。如此一來,擔任方案設計角色的GA,其創造力將受到限制,結果可能導致:決策者無法從缺乏創造力的GA所設計出的方案中找到滿意的解答。本文認為「找到決策者完整的目標集合」是IGA模型成功的一個重要關鍵。於是根據Keeney的價值焦點思考(Value Focused Thinking)概念,提出「如何找到決策者完整的目標集合」的做法,並將之整合至IGA中。

IGA的發展經歷幾個主要的里程,包括2005TAGAGINew Generation Computing編輯的IEC特刊。此特刊應該是IGA研究的一個里程碑,本著作被收錄於特刊中。隔年Tagagi與法國的 Lutton 共同籌劃第一屆以IGA為主題的 Workshop (The First European Workshop on Interactive Evolution and Humanized Computational Intelligence; EvoInteraction2006).,並與歐洲著名的演化式計算研討會 EuroGP 2006 & EvoCOP 2006 聯合在匈牙利的布達佩斯舉行。2007年在西班牙舉行第二屆Workshop (EvoInteraction 2007) 2009年在德國舉行第三屆Workshop (EvoInteraction 2009)。本人受邀連續擔任3屆議程委員。

大綱

The user fatigue problem in interactive evolutionary computation (IEC) is a complex and interesting issue. If the IEC search space is created from the experience or knowledge of domain experts rather than from users values, it causes two potential problems which lead to fatigue problems in IEC: 1) inefficiency and 2) boredom.

Therefore, we propose a customer values-based IEC model, solving the fatigue problem by avoiding the potential problems. A case study involving the design of mineral water bottles was used to verify the anti-fatigue capability of the users when using the proposed model. For comparison with the traditional domain knowledge-based model, we built two IEC systems, a customer values-based system and a traditional system, and conducted a user burden test and a system efficiency test over a two-week period. The results of both tests show that our proposed system performed better than the traditional system in designing mineral water bottles.

被引用情況

Kowaliw, T., Dorin, A., & McCormack, J. (2012). Promoting creative design in interactive evolutionary computation. IEEE transactions on evolutionary computation, 16(4), 523-536.

Chou, C., Kimbrough, S., Sullivan-Fedock, J., Woodard, C. J., & Murphy, F. H. (2012, July). Using interactive evolutionary computation (IEC) with validated surrogate fitness functions for redistricting. In Proceedings of the fourteenth international conference on Genetic and evolutionary computation (pp. 1071-1078). ACM.

Yan, S., Wanliang, W., & Xiaojian, L. (2010, September). An improved evaluation method for interactive genetic algorithms and its application in product design. In Bio-Inspired Computing: Theories and Applications (BIC-TA), 2010 IEEE Fifth International Conference on (pp. 840-843). IEEE.

Whigham, P. A., Aldridge, C., & De Lange, M. (2009, May). Constrained evolutionary art: Interactive flag design. In Evolutionary Computation, 2009. CEC’09. IEEE Congress on (pp. 2194-2200). IEEE.

Yang, H. F., & Lin, M. H. (2009). Innovative Chance Discovery–Extracting Customers’ Innovative Concept. Applications of Evolutionary Computing, 462-466.

Lewis, M. (2008). Evolutionary visual art and design. The Art of Artificial Evolution, 3-37.

Takagi, H., & Ohsaki, M. (2007). Interactive evolutionary computation-based hearing aid fitting.  IEEE Transactions on Evolutionary Computation, 11(3), 414-427.

Wang, L. H. (2007). A comparison of three fitness prediction strategies for interactive genetic algorithms. Journal of information science and engineering, 23(2), 605.

Hsu, F-C., Chi, T. H. & Pan, R. C. (2007). A Semi-Automatic Fitness Assignment Approach for Designing Products with Interactive Genetic Algorithms, Information Sciences, 210-214.

Hsu, F. C., & Hung, M. H. (2006). Practically applying interactive genetic algorithms to customers’ designs on a customizable c2c framework: entrusting select operations to IGA users. Applications of Evolutionary Computing, 575-585.

Hung, M. H., & Hsu, F. C. (2005). Accelerating Interactive Evolutionary Computation Convergence Pace by Using Over-sampling Strategy. Soft Computing as Transdisciplinary Science and Technology, 663-671.