IQL
An offline reinforcement learning algorithm that avoids overestimation by entirely refraining from querying the values of out-of-distribution actions. Along with CQL, it is a core method of offline RL.
Timeline (last 90 days)
Implicit Q-Learning (IQL) completely avoids querying the values of out-of-distribution actions.
Offline reinforcement learning penalizes agents for deviating from the dataset distribution by introducing constraint algorithms such as CQL and IQL, but at the cost of limited exploration capability.
All Facts (2)
Implicit Q-Learning (IQL) completely avoids querying the values of out-of-distribution actions.
50%UnverifiedOffline reinforcement learning penalizes agents for deviating from the dataset distribution by introducing constraint algorithms such as CQL and IQL, but at the cost of limited exploration capability.
50%