[This is preliminary documentation and is subject to change.]

This class implements the Action Centered Subsystem (ACS) in the Clarion Library

The ActionCenteredSubsystem..::..ActionCenteredSubsystemParameters type exposes the following members.

Properties

  Name Description
Public property A
The constant A used to factor in the MCS measure when calculating the probability or weight of for the bottom level during level selection
Public property B
The constant B used to factor in the MCS measure when calculating the probability or weight of for the bottom level during level selection
Public property BOTTOM_LEVEL_DECISION_TIME
The decision time for the bottom level of the ACS
Public property C3
The C3 constant used for variable level selection
Public property C4
The C4 constant used for variable level selection
Public property DEFAULT_ACTION_POTENTIAL
Specifies the default activation to set for the DO_NOTHING external action chunk (if no IDNs are eligible to be used) when calculating action potential
Public property DELETION_FREQUENCY
The frequency (in terms of # of learning steps) in which deletions by density are to be performed
Public property DISCOUNT
The match discount factor to be applied at the end of an episode
Public property EXTERNAL_ACTION_PROBABILITY
The probability of choosing an external action
Public property FIXED_BL_LEVEL_SELECTION_MEASURE
The fixed selection measure for the bottom level
Public property FIXED_FR_LEVEL_SELECTION_MEASURE
The fixed selection measure for the fixed rules
Public property FIXED_IRL_LEVEL_SELECTION_MEASURE
The fixed selection measure for the IRL rules
Public property FIXED_RER_LEVEL_SELECTION_MEASURE
The fixed selection measure for the RER rules
Public property GS_UPDATE_ACTION_PROBABILITY
The probability of choosing a goal structure update action
Public property LEVEL_SELECTION_METHOD
Specifies the method to use for level selection
Public property LEVEL_SELECTION_OPTION
Specifies the option to use for level selection
Public property LOCAL_EPISODIC_MEMORY_RETENTION_THRESHOLD
The maximum number of time steps (since the most recent) that local episodic memory should be kept before being discarded
Public property MAXIMUM_DECISION_TIME
The maximum decision-time for the ACS
Public property MCS_BL_SELECTION_MEASURE
The level selection measure specified by the the MCS (used for selecting the bottom level of the ACS)
Public property MCS_FR_SELECTION_MEASURE
The level selection measure specified by the the MCS (used for selecting the bottom level of the ACS)
Public property MCS_IRL_SELECTION_MEASURE
The level selection measure specified by the the MCS (used for selecting the bottom level of the ACS)
Public property MCS_MAXIMUM_DECISION_TIME
The maximum decision-time as reported by the MCS
Public property MCS_RER_SELECTION_MEASURE
The level selection measure specified by the the MCS (used for selecting the bottom level of the ACS)
Public property NACS_REASONING_ACTION_PROBABILITY
The probability of choosing an NACS reasoning action
Public property NACS_RETRIEVE_ACTION_PROBABILITY
The probability of choosing an NACS retrieve action
Public property PARAMETER_CHANGE_ACTION_PROBABILITY
The probability of choosing a parameter change action
Public property PERFORM_BL_LEARNING
Specifies whether or not bottom level learning should be performed
Public property PERFORM_DELETION_BY_DENSITY
Specifies whether or not intermittent rule deletion by density should be performed
Public property PERFORM_IRL_REFINEMENT
Specifies whether or not IRL rule refinement should be performed
Public property PERFORM_LEARNING
Specifies whether or not learning in general (including extraction, refinement, bottom-level learning, deletion by density, etc.) should be performed
Public property PERFORM_RER_REFINEMENT
Specifies whether or not RER rule refinement should be performed
Public property PERFORM_RULE_EXTRACTION
Specifies whether or not rule extraction should be performed
Public property PERFORM_TOP_DOWN_LEARNING
Specifies whether or not top-down learning should be performed
Public property PERSISTENCE
The persistence factor for the previous selection
Public property POSITIVE_MATCH_THRESHOLD
The threshold for the positive match criterion of the top and bottom level component collections
Public property SELECTION_TEMPERATURE
The temperature for stochastic selection
Public property SELECTION_THRESHOLD
The threshold for stochastic selection
Public property TOP_LEVEL_DECISION_TIME
The decision time for the top level of the ACS
Public property USE_ACTION_POTENTIAL
Specifies whether action potential should be calculated by the ACS during perception
Public property USE_ACTION_PROBABILITIES
Specifies whether action type selection (based on probability) should be performed
Public property VARIABLE_BL_BETA
The bottom level beta for variable level selection
Public property VARIABLE_FR_BETA
The fixed rule beta for variable level selection
Public property VARIABLE_IRL_BETA
The IRL beta for variable level selection
Public property VARIABLE_RER_BETA
The RER beta for variable level selection
Public property WM_UPDATE_ACTION_PROBABILITY
The probability of choosing a working memory update action
Public property WORKING_MEMORY_CAPACITY
The maximum number of slots for working memory (i.e., the maximum number of chunks that can be in working memory)

See Also