Uppercase are used for sets, and Greek letters represent parameters of the algorithms.

Set of states. | |

Individual states. Full views. | |

Number of states. | |

fd |
Set of feature detectors. |

fdfd |
Partial view of order . |

Set of actions of the robot. | |

Number of actions. | |

Set of elementary actions. | |

Number of motors of the robot. | |

Elementary action that assigns value to motor . | |

Partial command of order . | |

Action. Combination of elementary actions. Full command. | |

Partial rule composed by partial view and partial command . | |

The empty partial rule. | |

Composition of two partial rules. | |

Controller or set of partial rules. | |

Maximum number of elements of . | |

Subset of rules active at a given time step and at the previous one. | |

Active rules with a partial command in accordance with . | |

Expected value of the partial rule . | |

Expected error in the value estimation of the partial rule . | |

Average error in the value prediction. | |

Confidence index. | |

Confidence on the statistics of the partial rule . | |

Top value of the confidence. | |

Index where the confidence function reaches the value . | |

Error in the return prediction of the partial rule . | |

Relevance of rule . | |

Value interval of the partial rule . | |

Updating ratio for the statistics of the partial rule . | |

Learning rate. Top value of . | |

Number of times rule has been used. | |

Most relevant active partial rule w.r.t. action . | |

Most reliable value estimation for action . | |

Reward received after the execution of . | |

Discount factor. | |

Goodness of a given situation. | |

Value of executing action in given situation. | |

Number of new partial rules created at a time. | |

Redundancy threshold used for partial-rule elimination. |

Josep M Porta 2005-02-17