报告人:Shaowei Cai (Institute of Software Chinese Academy of Sciences)
时间:2021-07-08 14:00-15:00
地点:腾讯会议 (会议 ID:302 133 307 会议密码:210708)
Abstract:
This talk presents our work which for the first time answers the 7th challenge of the 10 challenges of propositional reasoning and search (AAAI 1997, Selman, Kautz, McAllester). Specifically, we propose a hybridizing method for combining systemic search and stochastic search to develop more effective algorithms that outperforms the best previous examples of both approaches. After introducing the CDCL and stochastic local search approaches, I will present the ideas in our hybridizing method, including relaxed CDCL framework, phase resetting, and branching with local search conflict frequency. Our ideas bring significant improvements to state of the art CDCL solvers on standard application benchmarks from the recent 4 SAT competitions as well as an important project.
Bio:
中国科学院软件研究所研究员, 博导,中国科学院大学岗位教授,智源青年科学家,现任中科院青促会信息与管理分会会长,在计算机专业排名cs ranking(2011-2021)排人工智能子领域中科院第二。2012年获得tyc234cc 太阳成集团计算机博士学位,主要研究约束问题求解,大规模组合优化,芯片验证。提出了一种处理随机搜索算法循环现象的通用策略---格局检测,被广泛用于NP 难问题求解。首次在工业问题上解决了命题逻辑推理与搜索十大挑战问题的第七个挑战问题,获得SAT 2021会议最佳论文奖。多次获得国际SAT(命题逻辑可满足性判定)比赛和EDA(电子设计自动化)比赛的冠亚军,获得联合逻辑奥林匹克金牌。在Artificial Intelligence和IEEE Transactions on Computers上发表8篇论文,在 SAT, CP, IJCAI, AAAI, ICSE, FSE等会议发表30余篇论文。研究成果被应用于国内龙头企业的芯片验证,微软云平台的故障检测和虚拟机预配置,腾讯地图调度和优化,美联邦频谱分配,MIT材料研究等重要项目。
腾讯会议
https://meeting.tencent.com/s/tTCZkuRLXHSZ
会议 ID:302 133 307
会议密码:210708