An interface for defining a database of state features that can be returned for any given input state or input state-action pair.
Many functions approximation techniques require a fixed feature vector to work and in many cases, using abstract features from the state attributes is useful.
A general interface for defining state value or Q-value function approximation and interacting with it via gradient descent methods.
A class that ties function approximation results to actions.
A class that associates an action with a set of state features.
A list associating a predicted value that was generated from a list of state features and the weights for those features.
This class holds the weight value for weights defined by a ValueFunctionApproximation class.
A class for associating a state feature identifier with a value of that state feature
A data structure for defining the gradient of the weights for a vector.