报告主题：蒙特卡洛树搜索方法在稀疏矩阵重排序问题的应用(An Efficient Single-player Monte Carlo Tree Search Method Based on Deep Learning and Element Importance for Sparse Matrix Reordering Problems)
报告人：戴彧虹 研究员 （中科院数学与系统科学研究院）
报告时间：2021年4月17日（周六） 14:30
报告地点：F309
邀请人：白延琴
报告摘要：The sparse matrix reordering problems is often used to produce the fewest new nonzero elements (called fill-ins) as possible as it can save computational cost and storage before applying direct methods to solve the large scale linear system. The sparse matrix reordering problems is NP-complete, so heuristic algorithms are usually used. This talk treats the sparse matrix reordering problems as a single-player game problem. Based on Deep Learning and element importance, an efficient single –player Monte Carlo tree search method is proposed to solve the sparse matrix reordering Problems.