Thanks to the recent achievements of Google DeepMind in 2013 and 2016, which succeeded at reaching so-called superhuman levels in Atari games and beat the world champion Go, RL has become very interesting in of the machine learning community. This renewed interest is also due to the advent of Deep Neural Networks (DNNs) as approximation functions, bringing the potential value of this type of algorithm to an even higher level. The algorithm that has gained the most interest in recent times is definitely Deep Q-Learning. The following section introduces the Deep Q-Learning algorithm and also discusses some optimization techniques to maximize its performance.
Deep Q neural networks
The Q-learning base algorithm can cause tremendous ...