Deep reinforcement learning

In the examples of the previous chapters, the estimates of the value function were made using a table, in which each box represents a state or a state–action pair. The use of a table to represent the value function allows the creation of simple algorithms and, if the environmental conditions are Markovian, allows to accurately estimate the value function because it assigns the expected return learned during policy iterations to every possible configuration from the environment. The use of the table, however, also leads to limitations; in fact, these methods are applicable only to environments with a reduced number of states and actions. The problem is not limited to the large amount of memory required to store the table, but also to the large amount of data and time required to estimate each state–action pair accurately. In other words...