Detection of Partial Task Graph Using Deep Learning
Tamura Taiga, 宗徳 甲斐 · 2019
Task scheduling is one of the optimization methods for parallel processing of programs. Task scheduling is intended to minimize execution time by allocating processing unit called task to appropriate computational resources. This method is considered to be impractical for large scale problems with the conventional search algorithms based on branch and bound method because of its computational complexity. One of the methods to solve this problem is to partially detect task graph and hierarchically conduct partial scheduling and complete scheduling. This can reduce computational complexity. But, this also causes another problem because the computation for detecting partial task graph itself is complicated. This research aims to solve this problem by using Deep Learning for detecting partial task graphs.