HARASE SHIN
Department / Course College of Science and Engineering Department of Mechanical Engineering
Title / Position Lecturer
Research Period 2014/04 ~ 2017/03
Research Topic Implementing Markov chain quasi-Monte Carlo methods
Research Type KAKENHI Research
Consignor Japan Society for the Promotion of Science
Research Program Type Grant-in-Aid for Young Scientists (B)
KAKENHI Grant No. 26730015
KAKENHI Category Statistical science
Responsibility Representative Researcher
Representative Person HARASE Shin
Details (1) We searched for higher-order quasi-Monte Carlo point sets with t-value and WAFOM both small. In addition, we searched for low-WAFOM point sets with the property that the number of points may be increased while retaining the existing points (so-called the extensibility). (2) We developed 64-bit maximally equidistributed F2-linear pseudorandom number generators. (3) We searched for a prototype version of Sobol' point sets with two-dimensional projections even better than the existing ones. We tested our point sets in financial engineering. (4) To construct quasi-Monte Carlo point sets for Markov chain Monte Carlo methods, we explored several kinds of algorithms. In particular, we noticed that short-period Tausworthe generators optimized in terms of regular continued fractions are probably hopeful.