A Survey of Deep Q-Networks used for Reinforcement Learning: State of the Art

Hafiz, A. M.Reinforcement learning (RL) is being intensely researched. The rewards lie with the goal of transitioning from human-supervised to machine-based automated decision making for real-world tasks. Many RL-based schemes are available. One such promising RL technique is deep reinforcement learning. This technique combines deep learning with RL. The deep networks having RL-based optimization goals are known as Deep Q-Networks after the well-known Q-learning algorithm. Many such variants of Deep Q-Networks are available, and more are being researched. In this paper, an attempt is made to give a gentle introduction to Deep Q-networks used for solving RL tasks as found in existing literature. The recent trends, major issues and future scope of DQNs are touched upon for benefit of the readers.