A modification of gradient policy in reinforcement learning procedure

摘要

The gradient of a scalar function is frequently used in various areas of mathematics. In informatics it can be used, for example, in the process of learning procedure of many control systems. The key observation is that gradient, if it is a non-zero vector, is a vector in the direction of greatest rate of the scalar function. In this contribution we show a method how to determine the direction(s) even if the gradient is zero vector. We show that this can be done with the knowledge which students have it their stage of study.