Deep Quality-Value (DQV) Learning
This is an incremental improvement for deep reinforcement learning practitioners, offering a potentially better synchronous temporal difference algorithm.
The paper tackles the problem of improving deep reinforcement learning efficiency by introducing Deep Quality-Value (DQV) Learning, which uses a value network to train a quality-value network, resulting in significantly faster and better learning than Deep Q-Learning and Double Deep Q-Learning on Atari games.
We introduce a novel Deep Reinforcement Learning (DRL) algorithm called Deep Quality-Value (DQV) Learning. DQV uses temporal-difference learning to train a Value neural network and uses this network for training a second Quality-value network that learns to estimate state-action values. We first test DQV's update rules with Multilayer Perceptrons as function approximators on two classic RL problems, and then extend DQV with the use of Deep Convolutional Neural Networks, `Experience Replay' and `Target Neural Networks' for tackling four games of the Atari Arcade Learning environment. Our results show that DQV learns significantly faster and better than Deep Q-Learning and Double Deep Q-Learning, suggesting that our algorithm can potentially be a better performing synchronous temporal difference algorithm than what is currently present in DRL.