学术报告第201219期

报告题目：1)   Evolutionary Squeaky Wheel Optimisation

                        2)   A pattern recognition based intelligent search  framework

报告人：Jingpeng  Li (李荆鹏博士)

报告时间：2012年12月19日下午15:30---16:30

报告地点：南一楼中311室

Abstract:

              1) Evolutionary Squeaky Wheel Optimisation

              Squeaky Wheel Optimization (SWO) is a relatively new search  method that has proved to be effective on many real-world problems. At each  iteration, SWO does a complete construction of a solution starting from the  empty assignment. Evolutionary SWO (ESWO) is a recent extension to SWO that is  designed to improve the intensification by keeping the good components of  solutions and only using SWO to reconstruct other poorer components of the  solution. In order to support the future study of such issues, we propose a  formal framework for the analysis of ESWO. The framework is based on Markov  chains, and the main novelty arises because ESWO moves through the space of  partial assignments. This makes it significantly different from the analyses  used in local search (such as simulated annealing) which only move through  complete assignments.

             2) A pattern recognition based intelligent search  framework

             Numerous papers based on various search methods across a  wide variety of applications have appeared in the literature. All of these  methods apply the following same approach to address the problems at hand: at  each iteration of the search, they first apply their search methods to generate  new solutions, then they calculate the objective values by taking some  constraints into account, and finally they use some strategies to determine the  acceptance or rejection of these solutions based upon the calculated objective  values. However, we suggest that calculating the exact objective value of every  resulting solution is not a must, particularly for highly constrained problems.  Furthermore, we believe that for newly-generated solutions, evaluating the  quality purely by their objective values is sometimes not the most efficient  approach. To address the above issues, we propose a pattern recognition-based  new framework towards the target of designing more intelligent and more flexible  search systems. The role of pattern recognition is to classify the quality of  resulting solutions, based on the solution structure rather than the solution  cost.

个人简介：

    李荆鵬1998年12月于304am永利集团数学系获硕士学位，2002年10月于英国利兹（Leeds）大学计算机学院获博士学位。2003年1月起在英国Bradford  大学信息学院任助理研究员一年，接着在诺丁汉（Nottingham）大学计算机学院任研究员7年，2010年8月被提升为永久职位的高级研究员，2011年9月回国来该大学宁波分校任教。在研究领域方面，他的方向属于人工智能和运筹学之间的交叉学科。具体而言，他从事于以下三方面进行研究：根据问题的不同属性设计不同的算法，来处理实际生活中经常要面临的一些具有挑战性的问题；发明更多的可运用在不同领域问题的更有效的新优化算法；对超启发式算法的有效性进行理论性的研究，该方向被公认为极为困难但意义重大。在个人研究成果方面，他博士毕业后先后在5个英国政府资助的大型科研项目中从事核心部分的研究，并作为第一作者或者通讯作者发表英文学术论文近30篇，其中绝大多数为本领域高SCI影响因子国际期刊文章。其发表的多项研究成果具有重要的理论和实际价值，成功地解决了欧洲﹑北美洲和中国在公共交通调度和医疗卫生规划方面的许多实际问题，取得了显著的社会效益。在资金申请方面，他目前主持宁波市自然基金1项，并在最近的5年内，与304am永利集团合作，作为前2名参与者实施了3个国家自然科学基金委员会资助的项目（其中2个在研1个已结题），并参与多个英国政府资助的大型科研项目（含1个1300万英镑和一个270万英镑）的立项报告书写和经费申请。在学术活动方面，他作为组织者长期活跃于国际重要学术交流活动的舞台，目前担任2家国际期刊的编委、10多个SCI国际权威期刊的审稿人以及20多国际学术会议的程序委员会委员。