A reinforcement learning algorithm that uses neural networks to estimate action values and reduces overestimation bias.