Actor-critic

Algorithm

Actor-critic algorithms are used in reinforcement learning and combine the advantages of policy-based algorithms like policy gradient estimation with value-based algorithms like Q-learning: one part of the algorithm, the actor, suggests an action to perform based on the current environment state parameters, and then a separate part of the algorithm, the critic, calculates the value function for this action with respect to the same environment state parameters. See also neural actor-critic.

alias
subtype
has functional building block
FBB_Behavioural modelling
has input data type
IDT_Vector of categorical variables IDT_Binary vector IDT_Vector of quantitative variables
has internal model
INM_Markov decision process
has output data type
ODT_Classification
has learning style
LST_Reinforcement
has parametricity
PRM_Parametric
has relevance
REL_Relevant
uses
sometimes supports
mathematically similar to