Efficient Algorithms for Graph Partitioning and Back Propagation in Numerical Applications
Hiroaki Yui · Institutional Repositories DataBase (IRDB) · 2016
Nowadays, the amount of data used in mapping, in mobile devices, and in genomic and astronomic sciences grows day by day in proportion to the spread of computers and mobiles equipped with a variety of sensors.To handle this extraordinary amount of information, it is important to analyze these data not only in the field of computer science but also in other science fields.However, algorithms for analyzing these data have led to considerable operating costs, and have yet to achieve complete accuracy.In data science, research to reduce the operating costs of data analysis using such algorithms, and to improve their accuracy, has already begun.Recently algorithms for analysis of these data are incurring smaller operating costs and offering improved accuracy.This thesis presents two new algorithms for analyzing data more efficiently.The first algorithm reduces operating cost when the algorithm successively repeats to solve a system of linear equations by substitution.The second algorithm improves the accuracy