ControlMechanism¶

Subclasses

Contents¶

Overview

ControlMechanisms and a Composition

Creating a ControlMechanism

Specifying OutputPorts to be monitored

Objective Mechanism

Specifying Parameters to Control

Structure

Input

Function

Output

Costs and Net Outcome

Execution

Examples

Class Reference

Overview¶

A ControlMechanism is a ModulatoryMechanism that modulates the value(s) of one or more Ports of other Mechanisms in the Composition to which it belongs. In general, a ControlMechanism is used to modulate the ParameterPort(s) of one or more Mechanisms, that determine the value(s) of the parameter(s) of the function(s) of those Mechanism(s). However, a ControlMechanism can also be used to modulate the function of InputPorts and/or OutputPort, much like a Gating Mechanism. A ControlMechanism’s function calculates a control_allocation: one or more values used as inputs for its control_signals. Its control_signals are ControlSignal OutputPorts that are used to modulate the parameters of other Mechanisms’ function (see Modulation for a more detailed description of how modulation operates). A ControlMechanism can be configured to monitor the outputs of other Mechanisms in order to determine its control_allocation, by specifying these in the monitor_for_control argument of its constructor, or in the monitor argument of an ObjectiveMechanism` assigned to its objective_mechanism argument (see Creating a ControlMechanism below). A ControlMechanism can also be assigned as the controller of a Composition, which has a special relation to that Composition: it generally executes either before or after all of the other Mechanisms in that Composition (see Controller Execution). The OutputPorts monitored by the ControlMechanism or its objective_mechanism, and the parameters it modulates can be listed using its show method.

ControlMechanisms and a Composition¶

A ControlMechanism can be assigned to a Composition and executed just like any other Mechanism. It can also be assigned as the controller of a Composition, that has a special relation to the Composition: it is used to control all of the parameters that have been specified for control in that Composition. A ControlMechanism can be the controller for only one Composition, and a Composition can have only one controller. When a ControlMechanism is assigned as the controller of a Composition (either in the Composition’s constructor, or using its add_controller method, the ControlMechanism assumes control over all of the parameters that have been specified for control for Components in the Composition. The Composition’s controller is executed either before or after all of the other Components in the Composition are executed, including any other ControlMechanisms that belong to it (see Controller Execution). A ControlMechanism can be assigned as the controller for a Composition by specifying it in the controller argument of the Composition’s constructor, or by using the Composition’s add_controller method. A Composition’s controller and its associated Components can be displayed using the Composition’s show_graph method with its show_control argument assigned as True.

Creating a ControlMechanism¶

A ControlMechanism is created by calling its constructor. When a ControlMechanism is created, the OutputPorts it monitors and the Ports it modulates can be specified in the montior_for_control and objective_mechanism arguments of its constructor, respectively. Each can be specified in several ways, as described below. If neither of those arguments is specified, then only the ControlMechanism is constructed, and its inputs and the parameters it modulates must be specified in some other way.

Specifying OutputPorts to be monitored¶

A ControlMechanism can be configured to monitor the output of other Mechanisms either directly (by receiving direct Projections from their OutputPorts), or by way of an ObjectiveMechanism that evaluates those outputs and passes the result to the ControlMechanism (see below for more detailed description). The following figures show an example of each:

_images/ControlMechanism_without_ObjectiveMechanism_fig.svg

_images/ControlMechanism_with_ObjectiveMechanism_fig.svg

Note that, in the figures above, the ControlProjections are designated with square “arrowheads”, and the ControlMechanisms are shown as septagons to indicate that their ControlProjections create a feedback loop (see Cycles and Feedback; also, see below regarding specification of a ControlMechanism and associated ObjectiveMechanism in a Composition’s add_linear_processing_pathway method).

Which configuration is used is determined by how the following arguments of the ControlMechanism’s constructor are specified (also see Examples):

monitor_for_control – a list of OutputPort specifications. If the objective_mechanism argument is not specified (or is False or None) then, when the ControlMechanism is added to a Composition, a MappingProjection is created from each OutputPort specified to InputPorts created on the ControlMechanism (see Input for details). If the objective_mechanism argument is specified, then the OutputPorts specified in monitor_for_control are assigned to the ObjectiveMechanism rather than the ControlMechanism itself (see Objective Mechanism for details).

objective_mechanism – if this is specfied in any way other than False or None (the default), then an ObjectiveMechanism is created that projects to the ControlMechanism and, when added to a Composition, is assigned Projections from all of the OutputPorts specified either in the monitor_for_control argument of the ControlMechanism’s constructor, or the monitor argument of the ObjectiveMechanism’s constructor (see Objective Mechanism for details). The objective_mechanism argument can be specified in any of the following ways:

False or None – no ObjectiveMechanism is created and, when the ControlMechanism is added to a Composition, Projections from the OutputPorts specified in the ControlMechanism’s monitor_for_control argument are sent directly to ControlMechanism (see specification of monitor_for_control argument).

True – an ObjectiveMechanism is created that projects to the ControlMechanism, and any OutputPorts specified in the ControlMechanism’s monitor_for_control argument are assigned to ObjectiveMechanism’s monitor argument instead (see Objective Mechanism for additional details).

a list of OutputPort specifications; an ObjectiveMechanism is created that projects to the ControlMechanism, and the list of OutputPorts specified, together with any specified in the ControlMechanism’s monitor_for_control argument, are assigned to the ObjectiveMechanism’s monitor argument (see Objective Mechanism for additional details).

a constructor for an ObjectiveMechanism – the specified ObjectiveMechanism is created, adding any OutputPorts specified in the ControlMechanism’s monitor_for_control argument to any specified in the ObjectiveMechanism’s monitor argument . This can be used to specify the function used by the ObjectiveMechanism to evaluate the OutputPorts monitored as well as how it weights those OutputPorts when they are evaluated (see below for additional details).

an existing ObjectiveMechanism – for any OutputPorts specified in the ControlMechanism’s monitor_for_control argument, an InputPort is added to the ObjectiveMechanism, along with MappingProjection to it from the specified OutputPort. This can be used to specify an ObjectiveMechanism with a custom function and weighting of the OutputPorts monitored (see below for additional details).

allow_probes – this argument allows values of Components of a nested Composition other than its OUTPUT Nodes to be specified in the monitor_for_control argument of the ControlMechanism’s constructor or, if objective_mechanism is specified, in the monitor argument of the ObjectiveMechanism’s constructor (see above). If the Composition’s allow_probes attribute is False, it is set to CONTROL, and only a ControlMechanism can receive projections from PROBE Nodes of a nested Composition (the current one as well as any others in the same Composition); if the Composition’s allow_probes attribute is True, then it is left that way, and any node within the Comopsition, including the ControlMechanism, can receive projections from PROBE Nodes (see Probes for additional details).

The OutputPorts monitored by a ControlMechanism or its objective_mechanism are listed in the ControlMechanism’s monitor_for_control attribute (and are the same as those listed in the monitor attribute of the objective_mechanism, if specified).

Note that the MappingProjections created by specification of a ControlMechanism’s monitor_for_control argument or the monitor argument in the constructor for an ObjectiveMechanism in the ControlMechanism’s objective_mechanism argument supersede any MappingProjections that would otherwise be created for them when included in the pathway argument of a Composition’s add_linear_processing_pathway method.

Objective Mechanism¶

If an ObjectiveMechanism is specified for a ControlMechanism (in the objective_mechanism argument of its constructor; also see Examples), it is assigned to the ControlMechanism’s objective_mechanism attribute, and a MappingProjection is created automatically that projects from the ObjectiveMechanism’s OUTCOME output_port to the OUTCOME input_port of the ControlMechanism.

The objective_mechanism is used to monitor the OutputPorts specified in the monitor_for_control argument of the ControlMechanism’s constructor, as well as any specified in the monitor argument of the ObjectiveMechanism’s constructor. Specifically, for each OutputPort specified in either place, an input_port is added to the ObjectiveMechanism. OutputPorts to be monitored (and corresponding input_ports) can be added to the objective_mechanism later, by using its add_to_monitor method. The set of OutputPorts monitored by the objective_mechanism are listed in its monitor attribute, as well as in the ControlMechanism’s monitor_for_control attribute.

When the ControlMechanism is added to a Composition, the objective_mechanism is also automatically added, and MappingProjectons are created from each of the OutputPorts that it monitors to its corresponding input_ports. When the Composition is run, the value(s) of the OutputPort(s) monitored are evaluated using the objective_mechanism's function, and the result is assigned to its OUTCOME output_port. That value is then passed to the ControlMechanism’s OUTCOME input_port, which is used by the ControlMechanism’s function to determine its control_allocation.

If a default ObjectiveMechanism is created by the ControlMechanism (i.e., when True or a list of OutputPorts is specified for the objective_mechanism argument of the constructor), then the ObjectiveMechanism is created with its standard default function (LinearCombination), but using PRODUCT (rather than the default, SUM) as the value of the function’s operation parameter. The result is that the objective_mechanism multiplies the values of the OutputPorts that it monitors, which it passes to the ControlMechanism. However, if the objective_mechanism is specified using either a constructor for, or an existing ObjectiveMechanism, then the defaults for the ObjectiveMechanism class – and any attributes explicitly specified in its construction – are used. In that case, if the LinearCombination with PRODUCT as its operation parameter are still desired, this must be explicitly specified. This is illustrated in the following examples.

The following example specifies a ControlMechanism that automatically constructs its objective_mechanism:

>>> from psyneulink import *
>>> my_ctl_mech = ControlMechanism(objective_mechanism=True)
>>> assert isinstance(my_ctl_mech.objective_mechanism.function, LinearCombination)
>>> assert my_ctl_mech.objective_mechanism.function.operation == PRODUCT

Notice that LinearCombination was assigned as the function of the objective_mechanism, and PRODUCT as its operation parameter.

By contrast, the following example explicitly specifies the objective_mechanism argument using a constructor for an ObjectiveMechanism:

>>> my_ctl_mech = ControlMechanism(objective_mechanism=ObjectiveMechanism())
>>> assert isinstance(my_ctl_mech.objective_mechanism.function, LinearCombination)
>>> assert my_ctl_mech.objective_mechanism.function.operation == SUM

In this case, the defaults for the ObjectiveMechanism’s class are used for its function, which is a LinearCombination function with SUM as its operation parameter.

Specifying the ControlMechanism’s objective_mechanism with a constructor also provides greater control over how ObjectiveMechanism evaluates the OutputPorts it monitors. In addition to specifying its function, the monitor_weights_and_exponents argument can be used to parameterize the relative contribution made by the monitored OutputPorts when they are evaluated by that function (see Examples).

Specifying Parameters to Control¶

This can be specified in either of two ways (see Examples in ControlSignal):

With a ControlMechanism itself

The parameters controlled by a ControlMechanism can be specified in the control argument of its constructor; the argument must be a specification for one more ControlSignals. The parameter to be controlled must belong to a Component in the same Composition as the ControlMechanism when it is added to the Composition, or an error will occur.

With a Parameter to be controlled by the controller of a Composition

Control can also be specified for a parameter where the parameter itself is specified, by including the specification of a ControlSignal, ControlProjection, or the keyword CONTROL in a tuple specification for the parameter. In this case, the specified parameter will be assigned for control by the controller of any Composition to which its Component belongs, when the Component is added to the Composition (see ControlMechanisms and a Composition). Conversely, when a ControlMechanism is assigned as the controller of a Composition, a ControlSignal is created and assigned to the ControlMechanism for every parameter of any Component in the Composition that has been specified for control.

In general, a ControlSignal is created for each parameter specified to be controlled by a ControlMechanism. These are a type of OutputPort that send a ControlProjection to the ParameterPort of the parameter to be controlled. All of the ControlSignals for a ControlMechanism are listed in its control_signals attribute, and all of its ControlProjections are listed in its control_projections attribute (see Examples).

Structure¶

Input¶

By default, a ControlMechanism has a single input_port named OUTCOME. If it has an objective_mechanism, then the OUTCOME input_port receives a single MappingProjection from the objective_mechanism's OUTCOME OutputPort (see Objective Mechanism for additional details). If the ControlMechanism has no objective_mechanism then, when it is added to a Composition, MappingProjections are created from the items specified in monitor_for_control directly to InputPorts on the ControlMechanism (see Specifying OutputPorts to be monitored for additional details). The number of InputPorts created, and how the items listed in monitor_for_control project to them is deterimined by the ControlMechanism’s outcome_input_ports_option. All of the Inports that receive Projections from those items, or the objective_mechanism if the ControlMechanism has one, are listed in its outcome_input_ports attribute, and their values in the outcome attribute. The latter is used as the input to the ControlMechanism’s function to determine its control_allocation.

Function¶

A ControlMechanism’s function uses its outcome attribute (the value of its OUTCOME InputPort) to generate one or more values used for ControlMechanism’s control_allocation. By default, its function is assigned the Identity Function, which takes a single value as its input and returns it as its output. That is then used as the ControlSignal’s control_allocation, which in turn is assigned as the allocation for all of the ControlMechanism’s control_signals. That is, by default, the ControlMechanism’s input is distributed as the allocation to each of its control_signals. This same behavior also occurs for any custom function assigned to a ControlMechanism that returns a 2d array with a single item in its outer dimension (axis 0). If a function is assigned that returns a 2d array with more than one item, and the number of those items (i.e., the length of the 2d array) is the same as the number of control_signals, then each item is assigned to a corresponding ControlSignal (in the order in which they are specified in the control_signals argument of the ControlMechanism’s constructor). However, these default behaviors can be modified by specifying that individual ControlSignals reference different items in the ControlMechanism’s value (see OutputPort variable).

Output¶

The OutputPorts of a ControlMechanism are ControlSignals (listed in its control_signals attribute). It has a ControlSignal for each parameter specified in the control argument of its constructor, that sends a ControlProjection to the ParameterPort for the corresponding parameter. The ControlSignals are listed in the control_signals attribute; since they are a type of OutputPort, they are also listed in the ControlMechanism’s output_ports attribute. The parameters modulated by a ControlMechanism’s ControlSignals can be displayed using its show method.

By default, each ControlSignal is assigned as its allocation the value of the corresponding item of the ControlMechanism’s control_allocation; however, subtypes of ControlMechanism may assign allocations differently. The default_allocation argument of the ControlMechanism’s constructor can be used to specify a default allocation for ControlSignals that have not been assigned their own default_allocation.

Warning

the default_variable argument of the ControlMechanism’s constructor is not used to specify a default allocation; the default_allocation argument must be used for that purpose. The default_variable argument should only be used to specify the format of the value of the ControlMechanism’s input_port (see `Mechanism_InputPorts for additional details), and its default input when it does not receive any Projections (see default_variable for additional details).

Note

While specifying default_allocation will take effect the first time the ControlMechanism is executed, a value specified for default_variable will not take effect (i.e., in determining control_allocation or values of the ControlSignals) until after the ControlMechahnism is executed, and thus will not influence the parameters it controls until the next round of execution (see Timing for a detailed description of the sequence of events during a Composition’s execution).

The allocation is used by each ControlSignal to determine its intensity, which is then assigned to the value of the ControlSignal’s ControlProjection. The value of the ControlProjection is used by the ParameterPort to which it projects to modify the value of the parameter it controls (see Modulation for a description of how a ControlSignal modulates the value of a parameter).

Costs and Net Outcome¶

A ControlMechanism’s control_signals are each associated with a set of costs, that are computed individually by each ControlSignal when they are executed by the ControlMechanism. The costs last computed by the control_signals are assigned to the ControlMechanism’s costs attribute. A ControlMechanism also has a set of methods – combine_costs, compute_reconfiguration_cost, and compute_net_outcome – that can be used to compute the combined costs of its control_signals, a reconfiguration_cost based on their change in value, and a net_outcome (the value of the ControlMechanism’s OUTCOME InputPort minus its combined costs), respectively (see Computation of Costs and Net_Outcome below for additional details). These methods are used by some subclasses of ControlMechanism (e.g., OptimizationControlMechanism) to compute their control_allocation. Each method is assigned a default function, but can be assigned a custom functions in a corrsponding argument of the ControlMechanism’s constructor (see links to attributes for details).

Reconfiguration Cost

A ControlMechanism’s reconfiguration_cost is distinct from the costs of the ControlMechanism’s ControlSignals, and in particular it is not the same as their adjustment_cost. The latter, if specified by a ControlSignal, is computed individually by that ControlSignal using its adjustment_cost_function, based on the change in its intensity from its last execution. In contrast, a ControlMechanism’s reconfiguration_cost is computed by its compute_reconfiguration_cost function, based on the change in its control_allocation ControlMechanism.control_allocation> from the last execution, that will be applied to all of its control_signals. By default, compute_reconfiguration_cost is assigned as the Distance function with the EUCLIDEAN metric).

Execution¶

A ControlMechanism is executed using the same sequence of actions as any Mechanism, with the following additions.

# FIX: 11/3/21: MODIFY TO INCLUDE POSSIBLITY OF MULTIPLE OUTCOME_INPUT_PORTS The ControlMechanism’s function takes as its input the value of its OUTCOME input_port (also contained in outcome). It uses that to determine the control_allocation, which specifies the value assigned to the allocation of each of its ControlSignals. Each ControlSignal uses that value to calculate its intensity, as well as its cost to modulate the value of the ParameterPort(s) for the parameter(s) it controls. Note that the modulated value of the parameter may not be used until the subsequent TRIAL of execution, if the ControlMechansim is not executed until after the Component to which the paramter belongs is executed (see note).

Computation of Costs and Net_Outcome¶

Once the ControlMechanism’s function has executed, if compute_reconfiguration_cost has been specified, then it is used to compute the reconfiguration_cost for its control_allocation (see above. After that, each of the ControlMechanism’s control_signals calculates its cost, based on its intensity. The ControlMechanism then combines these with the reconfiguration_cost using its combine_costs function, and the result is assigned to the costs attribute. Finally, the ControlMechanism uses this, together with its outcome attribute, to compute a net_outcome using its compute_net_outcome function. This is used by some subclasses of ControlMechanism (e.g., OptimizationControlMechanism) to compute its control_allocation for the next TRIAL of execution.

Execution as Controller of a Composition¶

if a ControlMechanism is assigned as the controller of a `Composition, then it is executed either before or after all of the other Mechanisms executed in a TRIAL for that Composition, depending on the value assigned to the Composition’s controller_mode attribute (see Controller Execution). If a ControlMechanism is added to a Composition for which it is not a controller, then it executes in the same way as any Mechanism, based on its place in the Composition’s graph. Because ControlProjections are likely to introduce cycles (recurrent connection loops) in the graph, the effects of a ControlMechanism and its projections will generally not be applied in the first TRIAL (see Cycles and Feedback for configuring the initialization of feedback loops in a Composition; also see Scheduler for a description of additional ways in which a ControlMechanism and its dependents can be scheduled to execute).

Examples

The examples below focus on the specification of the objective_mechanism for a ControlMechanism. See Control Signal Examples for examples of how to specify the ControlSignals for a ControlMechanism.

The following example creates a ControlMechanism by specifying its objective_mechanism using a constructor that specifies the OutputPorts to be monitored by its objective_mechanism and the function used to evaluate these:

>>> my_mech_A = ProcessingMechanism(name="Mech A")
>>> my_DDM = DDM(name="My DDM")
>>> my_mech_B = ProcessingMechanism(function=Logistic,
...                                 name="Mech B")

>>> my_control_mech = ControlMechanism(
...                          objective_mechanism=ObjectiveMechanism(monitor=[(my_mech_A, 2, 1),
...                                                                           my_DDM.output_ports[RESPONSE_TIME]],
...                                                                 name="Objective Mechanism"),
...                          function=LinearCombination(operation=PRODUCT),
...                          control_signals=[(THRESHOLD, my_DDM),
...                                           (GAIN, my_mech_B)],
...                          name="My Control Mech")

This creates an ObjectiveMechanism for the ControlMechanism that monitors the primary OutputPort of my_mech_A and the RESPONSE_TIME OutputPort of my_DDM; its function first multiplies the former by 2, then takes product of their values and passes the result as the input to the ControlMechanism. The ControlMechanism’s function uses this value to determine the allocation for its ControlSignals, that control the value of the threshold parameter of the DriftDiffusionAnalytical Function for my_DDM and the gain parameter of the Logistic Function for my_transfer_mech_B.

The following example specifies the same set of OutputPorts for the ObjectiveMechanism, by assigning them directly to the objective_mechanism argument:

>>> my_control_mech = ControlMechanism(
...                             objective_mechanism=[(my_mech_A, 2, 1),
...                                                  my_DDM.output_ports[RESPONSE_TIME]],
...                             control_signals=[(THRESHOLD, my_DDM),
...                                              (GAIN, my_mech_B)])

Note that, while this form is more succinct, it precludes specifying the ObjectiveMechanism’s function. Therefore, the values of the monitored OutputPorts will be added (the default) rather than multiplied.

The ObjectiveMechanism can also be created on its own, and then referenced in the constructor for the ControlMechanism:

>>> my_obj_mech = ObjectiveMechanism(monitored_output_ports=[(my_mech_A, 2, 1),
...                                                            my_DDM.output_ports[RESPONSE_TIME]],
...                                      function=LinearCombination(operation=PRODUCT))

>>> my_control_mech = ControlMechanism(
...                        objective_mechanism=my_obj_mech,
...                        control_signals=[(THRESHOLD, my_DDM),
...                                         (GAIN, my_mech_B)])

Here, as in the first example, the constructor for the ObjectiveMechanism can be used to specify its function, as well as the OutputPort that it monitors.

Class Reference¶

class psyneulink.core.components.mechanisms.modulatory.control.controlmechanism.ControlMechanism(default_variable=None, input_shapes=None, monitor_for_control=None, objective_mechanism=None, allow_probes=False, outcome_input_ports_option=None, function=None, default_allocation=None, control=None, modulation=None, combine_costs=None, compute_reconfiguration_cost=None, compute_net_outcome=None, params=None, name=None, prefs=None, **kwargs)¶

Subclass of ModulatoryMechanism that modulates the parameter(s) of one or more Component(s). See Mechanism for additional arguments and attributes.

Parameters:

monitor_for_control (List[OutputPort or Mechanism] : default None) – specifies the OutputPorts to be monitored by the ObjectiveMechanism, if specified in the objective_mechanism argument (see Objective Mechanism), or directly by the ControlMechanism itself if an objective_mechanism is not specified. If any specification is a Mechanism (rather than its OutputPort), its primary OutputPort is used (see Specifying OutputPorts to be monitored for additional details).
objective_mechanism (ObjectiveMechanism or List[OutputPort specification] : default None) – specifies either an ObjectiveMechanism to use for the ControlMechanism, or a list of the OutputPorts it should monitor; if a list of OutputPort specifications is used, a default ObjectiveMechanism is created and the list is passed to its monitor argument, along with any OutputPorts specified in the ControlMechanism’s monitor_for_control argument.
allow_probes (bool : default False) – specifies whether Components of a nested Composition that are not OUTPUT <NodeRole.OUTPUT>` Nodes of that Composition can be specified as items in the ControlMechanism’s monitor_for_control argument or, if objective_mechanism is specified, in the ObjectiveMechanism’s monitor argument (see allow_probes for additional information).
outcome_input_ports_option (COMBINE, CONCATENATE, SEPARATE : default SEPARATE) – if objective_mechanism is not specified, this specifies whether MappingProjections from items specified in monitor_for_control are each assigned their own InputPort (SEPARATE) or to a single OUTCOME InputPort (CONCATENATE, COMBINE); (see outcome_input_ports_option for additional details.
function (TransferFunction : default Linear(slope=1, intercept=0)) – specifies function used to combine values of monitored OutputPorts.
control (ControlSignal specification or list[ControlSignal specification, ...]) – specifies the parameters to be controlled by the ControlMechanism; a ControlSignal is created for each (see Specifying ControlSignals for details of specification).
modulation (str : MULTIPLICATIVE) – specifies the default form of modulation used by the ControlMechanism’s ControlSignals, unless they are individually specified.
combine_costs (Function, function or method : default np.sum) – specifies function used to combine the cost of the ControlMechanism’s control_signals; must take a list or 1d array of scalar values as its argument and return a list or array with a single scalar value.
compute_reconfiguration_cost (Function, function or method : default None) – specifies function used to compute the ControlMechanism’s reconfiguration_cost; must take a list or 2d array containing two lists or 1d arrays, both with the same shape as the ControlMechanism’s control_allocation attribute, and return a scalar value.
compute_net_outcome (Function, function or method : default lambda outcome, cost: outcome-cost) – function used to combine the values of its outcome and costs attributes; must take two 1d arrays (outcome and cost) with scalar values as its arguments and return an array with a single scalar value.

monitor_for_control¶

each item is an OutputPort monitored by the ControlMechanism or its objective_mechanism if that is specified (see Specifying OutputPorts to be monitored); in the latter case, the list returned is ObjectiveMechanism’s monitor attribute.

Type:: List[OutputPort]

objective_mechanism¶

ObjectiveMechanism that monitors and evaluates the values specified in the ControlMechanism’s objective_mechanism argument, and transmits the result to the ControlMechanism’s OUTCOME input_port.

Type:: ObjectiveMechanism

allow_probes¶

indicates status of the allow_probes attribute of the Composition to which the ControlMechanism belongs. If False, items specified in the monitor_for_control are all OUTPUT Nodes of that Composition. If True, they may be INPUT or INTERNAL Nodes of nested Composition (see allow probes and Composition_Probes for additional information).

Type:: bool

outcome_input_ports_option¶

determines how items specified in monitor_for_control project to the ControlMechanism if no objective_mechanism is specified. If SEPARATE is specified (the default), the Projection from each item specified in monitor_for_control is assigned its own InputPort. All of the InputPorts are assigned to a list in the ControlMechanism’s outcome_input_ports attribute. If CONCATENATE or COMBINE is specified, all of the projections are assigned to a single InputPort, named OUTCOME. If COMBINE is specified, the OUTCOME InputPort is assigned LinearCombination as its function, which sums the values of the projections to it (all of which must have the same dimension), to produce a single array (this is the default behavior for multiple Projections to a single InputPort; see InputPort function). If CONCATENATE is specified, the OUTCOME InputPort is assigned Concatenate as its function, which concatenates the values of its Projections into a single array of length equal to the sum of their lengths (which need not be the same). In both cases, the OUTCOME InputPort is assigned as the only item in the list of outcome_input_ports.

Type:: , SEPARATE, COMBINE, or CONCATENATE

monitored_output_ports_weights_and_exponents¶

each tuple in the list contains the weight and exponent associated with a corresponding OutputPort specified in monitor_for_control; if objective_mechanism is specified, these are the same as those in the ObjectiveMechanism’s monitor_weights_and_exponents attribute, and are used by the ObjectiveMechanism’s function to parametrize the contribution made to its output by each of the values that it monitors (see ObjectiveMechanism Function).

Type:: List[Tuple(float, float)]

outcome_input_ports¶

list of the ControlMechanism’s InputPorts that receive Projections from either is objective_mechanism (in which case the list contains only the ControlMechanism’s OUTCOME InputPort), or the OutputPorts of the items listed in its monitor_for_control attribute.

Type:: ContentAddressableList

outcome¶

an array containing the value of each of the ControlMechanism’s outcome_input_ports.

Type:: 1d array

function¶

determines how the values of the OutputPorts specified in the monitor_for_control argument of the ControlMechanism’s constructor are used to generate its control_allocation.

Type:: TransferFunction : default Linear(slope=1, intercept=0)

default_allocation¶

determines the default_allocation of any control_signals for which the default_allocation is not specified in its constructor; if it is None (not specified) then the ControlSignal’s parameters.allocation.default_value is used. See documentation for default_allocation argument of ControlSignal constructor for additional details.

Type:: number, list or 1d array

control_allocation¶

each item is the value assigned as the allocation for the corresponding ControlSignal listed in the control_signals attribute (that is, it is a list of the values of the variable attributes of the ControlMechanism’s ControlSignals).

Type:: 2d array

control_signals¶

list of the ControlSignals for the ControlMechanism, including any inherited from a Composition for which it is a controller (same as ControlMechanism’s output_ports attribute); each sends a ControlProjection to the ParameterPort for the parameter it controls.

Type:: ContentAddressableList[ControlSignal]

compute_reconfiguration_cost¶

function used to compute the ControlMechanism’s reconfiguration_cost; result is a scalar value representing the difference — defined by the function — between the values of the ControlMechanism’s current and last control_alloction, that can be accessed by reconfiguration_cost attribute.

Type:: Function, function or method

reconfiguration_cost¶

result of compute_reconfiguration_cost function, that computes the difference between the values of the ControlMechanism’s current and last control_alloction; value is None and is ignored if compute_reconfiguration_cost has not been specified.

A ControlMechanism’s reconfiguration_cost is not the same as the adjustment_cost of its ControlSignals (see Reconfiguration Cost for additional details).

Type:: scalar

costs¶

current costs for the ControlMechanism’s control_signals, computed for each using its compute_costs method.

Type:: list

combine_costs¶

function used to combine the cost of its control_signals; result is an array with a scalar value that can be accessed by combined_costs.

Note

This function is distinct from the combine_costs_function of a ControlSignal. The latter combines the different costs for an individual ControlSignal to yield its overall cost; the ControlMechanism’s combine_costs function combines those costs for its control_signals.

Type:: Function, function or method

combined_costs¶

result of the ControlMechanism’s combine_costs function.

Type:: 1d array

compute_net_outcome¶

function used to combine the values of its outcome and costs attributes; result is an array with a scalar value that can be accessed by the the net_outcome attribute.

Type:: Function, function or method

net_outcome¶

result of the ControlMechanism’s compute_net_outcome function.

Type:: 1d array

control_projections¶

list of ControlProjections that project from the ControlMechanism’s control_signals.

Type:: List[ControlProjection]

modulation¶

the default form of modulation used by the ControlMechanism’s ControlSignals, unless they are individually specified.

Type:: str

outputPortTypes¶: alias of ControlSignal

_validate_params(request_set, target_set=None, context=None)¶: Validate monitor_for_control, objective_mechanism, CONTROL_SIGNALS and GATING_SIGNALS

_instantiate_objective_mechanism(input_ports=None, context=None)¶

# FIX: ??THIS SHOULD BE IN OR MOVED TO ObjectiveMechanism Assign InputPort to ObjectiveMechanism for each OutputPort to be monitored;

uses _instantiate_monitoring_input_port and _instantiate_control_mechanism_input_port to do so. For each item in self.monitored_output_ports: - if it is a OutputPort, call _instantiate_monitoring_input_port() - if it is a Mechanism, call _instantiate_monitoring_input_port for relevant Mechanism_Base.output_ports

(determined by whether it is a TERMINAL Mechanism and/or MonitoredOutputPortsOption specification)

each InputPort is assigned a name with the following format:
‘<name of Mechanism that owns the monitoredOutputPort>_<name of monitoredOutputPort>_Monitor’

Notes: * self.monitored_output_ports is a list, each item of which is a Mechanism_Base.output_port from which a

Projection will be instantiated to a corresponding InputPort of the ControlMechanism

self.input_ports is the usual ordered dict of ports,
each of which receives a Projection from a corresponding OutputPort in self.monitored_output_ports

_instantiate_input_ports(input_ports=None, context=None)¶

Instantiate input_ports for items being monitored and evaluated, and ObjectiveMechanism if specified

If objective_mechanism is specified:

instantiate ObjectiveMechanism, which also instantiates an OUTCOME InputPort and a MappingProjection to it from the ObjectiveMechanisms OUTCOME OutputPort

If monitor_for_control is specified:

it is used to construct an InputPort from each sender specified in it, and a corresponding MappingProjection from the sender to that InputPort;
each InputPort is named using an uppercase version of the sender’s name

If nothing is specified, a default OUTCOME InputPort is instantiated with no projections to it

_parse_monitor_for_control_input_ports(context)¶

Get outcome_input_port specification dictionaries for items specified in monitor_for_control.

Note: leave Projections unspecified, as they need to be added to self.aux_components: for validation and activation by Composition

Return port specification dictionaries (without Projections to them specified), their value sizes, and monitored ports (to be used as Projection specifications by _instantiate_input_ports)

_validate_monitor_for_control(nodes)¶: Ensure all of the Components being monitored for control are in the Composition being controlled If monitor_for_control is specified as an ObjectiveMechanism, warn and move to objective_mecahnism arg

_instantiate_output_ports(context=None)¶

Call Port._instantiate_output_ports to instantiate orderedDict of OutputPort(s)

This is a stub, implemented to allow Mechanism subclasses to override _instantiate_output_ports: or process InputPorts before and/or after call to _instantiate_output_ports

_instantiate_control_signals(context)¶: Subclasses can override for class-specific implementation (see OptimizationControlMechanism for example)

_instantiate_control_signal(control_signal, context=None)¶

Parse and instantiate ControlSignal (or subclass relevant to ControlMechanism subclass)

Temporarily assign variable to default allocation value to avoid chicken-and-egg problem:: value, output_ports and control_signals haven’t been expanded yet to accomodate the new ControlSignal; reassign control_signal.variable to actual OWNER_VALUE below, once value has been expanded

_instantiate_control_signal_type(control_signal_spec, context)¶: Instantiate actual ControlSignal, or subclass if overridden

_set_mechanism_value(context)¶: Set Mechanism’s value. By default, use value returned by ControlMechanism’s function. Can be overridden by a subclass if it determines its value in some other way (see OCM for example). Note: this is used to determine the number of ControlSignals

_check_for_duplicates(control_signal, control_signals, context)¶

Check that control_signal is not a duplicate of one already instantiated for the ControlMechanism

Can happen if control of parameter is specified in constructor for a Mechanism: and also in the ControlMechanism’s control arg

control_signals arg passed in to allow override by subclasses

Warn if control_signal shares any ControlProjections with others in control_signals. Warn if control_signal is a duplicate of any in control_signals.

Return True if control_signal is a duplicate

show()¶: Display the OutputPorts monitored by ControlMechanism’s objective_mechanism and the parameters modulated by its control_signals.

add_to_monitor(monitor_specs, context=None)¶

Instantiate OutputPorts to be monitored by ControlMechanism’s objective_mechanism.

monitored_output_ports can be any of the following:

If any item is a Mechanism, its primary OutputPort is used. OutputPorts must belong to Mechanisms in the same System as the ControlMechanism.

_activate_projections_for_compositions(composition=None, context=None)¶: Activate eligible Projections to or from Nodes in Composition. If Projection is to or from a node NOT (yet) in the Composition, assign it the node’s aux_components attribute but do not activate it.

property _dependent_components¶: Returns: Components that must have values in a given Context for this Component to execute in that Context

exception psyneulink.core.components.mechanisms.modulatory.control.controlmechanism.ControlMechanismError(message, data=None)¶