DDQN Epsilon-greedy Benchmark
Proposed by DeepMind in 2015, Double-DQN is an extension of DQN to prevent value overestimation. It uses a separate target network to estimate the target value, and the two networks are rotated periodically.
Proposed by DeepMind in 2015, Double-DQN is an extension of DQN to prevent value overestimation. It uses a separate target network to estimate the target value, and the two networks are rotated periodically.