参会心得 学生：徐庆征 2009-6-19

提纲 会议基本信息 精彩报告 几点体会

会议基本信息 会议名称：2009 World Summit on Genetic and Evolutionary Computation 会议地点： 主办方： 共同组织者：

论文数量 Submissions: 372 Full paper:153 Post paper: 126

会议规模 Keynote: 1 Tutorial: 14 Intro: 6 Special: 3 Advanced: 5 Session: 28 Scheduled Paper Session: 27 Special Poster Session: 1

投稿过程 提交日期：2008-12-5 录用日期：2009-2-24 提交正式稿日期：2009-3-3 注册日期：2009-3-11 会议日期：2009-6-12至2009-6-14

会议经历 开幕式 Keynote Tutorial: 6 Paper Presentation: 16 做报告

精彩报告1 Keynote: Practical &Philosophical Reflections on a Life in Genetic Algorithms Tutorial: Fast, Effective GA’s for Large, Hard Problems Author: David E. Goldberg

Lessons Learning to Ask Learning to Label Learning to Decompose Learning to Model

Lesson 1: Learning to Ask In 1984 had many questions about how GAs work, when they fail? Wasn’t experienced in asking good framing questions. Key problem: Using GAs to solve engineering problems, but GAs weren’t engineered well. Socrates (470-399 BCE)

What’s a Good Question? Socrates asked variety of questions. What is truth? What is courage? More often the critic. Rarely gave answers. In creative enterprises, many good questions are framing questions: – Get at heart of the issue. – Help define the problem or elicit definition. – Sometimes cause problem to be represented in novel way or from unusual or creative perspective. Fundamental importance of dialectic. Creative process of asking and answering questions.

Lesson 2: Learning to Label Terms Really Do Matter Terms gather thoughts under consistent rubrics. Can be part of larger taxonomy. Defines attention areas. Can have influence on how others think. Catchy or sticky terms propagate virally.

Lesson 3: Learning to Decompose December 17, 1903: The Most Famous Moment in Aviation History

The Wright Brothers’ Secret Functional decomposition. Three subproblems: Stability: wing-warping plus elevator in 1899 glider model. 1902 glider had three-axis active control. Lift and Drag: wing shape improved on Lilenthal’s through air tunnel experiments. Propulsion: rotary wing with forward lift is a propeller.

Effective Theory in GA Design Many GAs don’t scale & much GA theory inapplicable. Need design theory that works: Understand building blocks (BBs), notions or subideas. Ensure BB supply. Ensure BB growth. Control BB speed. Ensure good BB decisions. Ensure good BB mixing (exchange). Know BB challengers. Can use theory to design scalable & efficient GAs.

Lesson 4: Learning to Model A Model of Models Error, ε Cost of Modeling, C Engineer/Inventor Scientist/Mathematician

A Life in Genetic Algorithms Events Bumped into GAs by accident. Joined field at time of growth. Fluids training as disciplinary grounding in complexity. Wrote a book I was told not to write. Became philosophical in a action-oriented field. Took on reform effort not admired by peers. Lessons? Important things can be random. Opportunity is knocking? Will you answer the door? Being appropriately different can be beneficial. Authority figures are not necessarily right or wise. Exploring the unexplored can yield interesting insights. Sometimes important jobs are not valued by others.

精彩报告2 Tutorial: Introduction to Genetic Algorithms Tutorial: Introduction to Genetic algorithm Theory and Practice Erik Goodman Darrell Whitley

精彩报告3 Tutorial: A Unified Framwork for Evolutionary Computation Author: Ken De Jong

Historical Roots Evolution Strategies (ESs) －developed by Rechenberg, Schwefel, etc. in 1960s Evolutionary Programming (EP) －developed by Fogel in 1960s Genetic Algorithms (GAs) －developed by Holland in 1960s

Present Status wide variety of evolutionary algorithms (EAs) wide variety of applications – optimization – search – learning, adaptation well-developed analysis – theoretical – experimental

Viewpoint Develop a general framework that: – Helps one compare and contrast approaches. – Encourages crossbreeding. – Facilitates intelligent design choices.

An EA Template Basic elements: 1. Randomly generate an initial population. 2. Do until some stopping criteria is met: Select individuals to be parents (biased by fitness). Produce offspring. Select individuals to die (biased by fitness). End Do. 3. Return a result. Basic elements: – a population of “individuals” – a notion of “fitness” – a birth/death cycle biased by fitness – a notion of “inheritance”

New Developments and Directions Exploiting parallelism: – coarsely grained network models – finely grained diffusion models Co-evolutionary models: – competitive co-evolution Exploiting Morphogenesis: – sophisticated genotype --> phenotype mappings – evolve plans for building complex objects rather than the objects themselves.

New Developments and Directions Self-adaptive EAs: – dynamically adapt to problem characteristics: – goal: robust “black box” optimizer Hybrid Systems: – combine EAs with other techniques Time-varying environments: – fitness landscape changes during evolution – goal: adaptation, tracking – standard optimization-oriented EAs not wellsuited for this.

New Developments and Directions Agent-oriented problems: – individuals more autonomous, active – fitness a function of other agents and environment-altering actions – standard optimization-oriented EAs not wellsuited for this.

精彩报告4 Paper Presentation: A Population Based Hybrid Meta-heuristic for the Uncapacitated Facility Location Problem Author: Wayne Pullan

精彩报告5 Paper Presentation: A Global Optimization Based on Physicomimetics Authors: Li-Ping Xie, Jian-Chao Zeng

几点体会 国际化 -注册 -论文集 -食宿 -日程安排 高水平 -“牛人” -优秀论文

几点体会 理论成果丰富，应用领域日益扩大 -美国领先，中国积极 加强英语学习，尤其是听、说能力 -锻炼论文写作能力，注意学术规范 积极开展对外合作，努力完善信息共享平台

几点体会 鼓励研究生参与学术交流 国内国际会议 教育部研究生教育创新计划 －全国研究生暑期学校 －全国博士生学术会议 －全国博士生学术论坛 －博士生国内访学 －地方研究生教育创新计划项目-西部地区研究生精品课程大 讲堂

更多信息 会议信息 http://www.sigevo.org/gec-summit-2009/ 论文信息 www.acm.org