
Demystifying Double DQNs: the What, the Why and the How
·5 min read
(from a web-based version here) The What It is a off-policy, model-free and value-based method, the same family as DQNs but with a subtle difference for optimization purposes! What's the optimization
(from a web-based version here) The What It is a off-policy, model-free and value-based method, the same family as DQNs but with a subtle difference for optimization purposes! What's the optimization and why was it necessary?! Well, with the previous DQN, a problem that it did not addressed, is that of overestimation of Q values from the Target Q Network, which leads to an overestimation bias Bias , in statistics, is the systematic difference between an estimator’s expected value and the true va
16
0
1
0
