Double Q-learning can learn from off-policy data. But it's fairly tricky to get all the tuning parameters right. It's a good choice if you can start up 100 or m

Double Q-learning can learn from off-policy data. But it's fairly tricky to get all the tuning parameters right. It's a good choice if you can start up 100 or more runs with a range of parameters and pick the one that worked best.