rcognita.critics

This module containing critics, which are integrated in controllers (agents).

Remarks:

  • All vectors are treated as of type [n,]

  • All buffers are treated as of type [L, n] where each row is a vector

  • Buffers are updated from bottom to top

Classes

Critic(system_dim_input, system_dim_output, …)

Critic base class.

CriticCALF(*args[, safe_decay_rate, …])

CriticOfActionObservation(system_dim_input, …)

This is the class of critics that are represented as functions of observation only.

CriticOfObservation(system_dim_input, …)

This is the class of critics that are represented as functions of observation only.

CriticTabularPI(*args[, tolerance, …])

CriticTabularVI(dim_state_space, …[, …])

Critic for tabular agents.

CriticTrivial(running_objective, *args[, …])

This is a dummy to calculate outcome (accumulated running objective).