Learning in proximity

In most articles about machine learning, particularly in reinforcement learning, a learning system interacts with an unknown environment and tries to improve its performance (receiving fewer penalties) according to feedback from environment as reinforcement signals. In this paper we show how a learning system can learn to interact with environment from an experienced agent (experienced learning system). Simulation results show that when a learning system learns from an experienced agent before exposure, it can interact better than a same raw learning system and receive fewer penalties from the environment.