Table of Contents

IntegratorMechanism

Contents

Overview

An IntegratorMechanism integrates its input, possibly based on its prior values. The input can be a single scalar value or an array of scalars (list or 1d np.array). If it is a list or array, then each value is independently integrated. The default function (IntegratorFunction) can be parametrized to implement either a simple increment rate, additive accumulator, or an (exponentially weighted) time-averaging of its input. It can also be assigned a custom function.

Creating an IntegratorMechanism

An IntegratorMechanism can be created directly by calling its constructor, or using the mechanism command and specifying INTEGRATOR_MECHANISM as its mech_spec argument. Its function is specified in the function argument, which can be parametrized by calling its constructor with parameter values:

>>> import psyneulink as pnl
>>> my_time_averaging_mechanism = pnl.IntegratorMechanism(function=pnl.AdaptiveIntegrator(rate=0.5))

The default_variable argument specifies the format of its input (i.e., whether it is a single scalar or an array), as well as the value to use if none is provided when Mechanism is executed. Alternatively, the input_shapes argument can be used to specify the length of the array, in which case it will be initialized with all zeros.

Structure

An IntegratorMechanism has a single InputPort, the value of which is used as the variable for its function. The default for function is AdaptiveIntegrator(rate=0.5). However, a custom function can also be specified, so long as it takes a numeric value, or a list or np.ndarray of numeric values as its input, and returns a value of the same type and format. The Mechanism has a single OutputPort, the value of which is assigned the result of the call to the Mechanism’s function.

Execution

When an IntegratorMechanism is executed, it carries out the specified integration, and assigns the result to the value of its primary OutputPort. For the default function (IntegratorFunction), if the value specified for default_variable is a list or array, or input_shapes is greater than 1, each element of the array is independently integrated. If its rate parameter is a single value, that rate is used for integrating each element. If the rate parameter is a list or array, then each element is used as the rate for the corresponding element of the input (in this case, rate must be the same length as the value specified for default_variable or input_shapes). Integration can be reset to the value of its functions initializer by setting its `reset parameter to a non-zero value, as described below.

Resetting the IntegratorMechanism

An IntegatorMechanism has a modulable reset parameter that can be used to reset its value to the value of its functions initializer. This also clears the value history, thus effectively setting the previous_value of its function to None.

The reset parameter can be used to reset the IntegratorMechanism under the control of a ControlMechanism. This simplest way to do this is to specify the reset parameter of the IntgeratorMechanism in the control argument of the ControlMechanism’s constructor, and to specify OVERRIDE in its modulation argument, as in the following example:

>>> my_integrator = IntegratorMechanism()
>>> ctl_mech = pnl.ControlMechanism(modulation=pnl.OVERRIDE, control=(pnl.RESET, my_integrator))

In this case, any non-zero value of the ControlMechanism’s ControlSignal will reset the IntegratorMechanism. OVERRIDE must be used as its modulation parameter (instead of its default value of MULTIPLICATIVE), so that the value of the ControlMechanism’s ControlSignal is assigned directly to the IntegratorMechanism’s reset parameter (otherwise, since the default of the reset parameter is 0, the ControlSignal’s value has no effect). An alternative is to specify the reset_default agument in the IntegratorMechanism constructor with a non-zero value, and while allowing the ControlMechanism to use its default value for modulation (i.e., MULTIPLICATIVE):

>>> my_integrator = IntegratorMechanism(reset_default=1)
>>> ctl_mech = pnl.ControlMechanism(control=(pnl.RESET, my_integrator))

In this case, a ControlSignal with a zero value suppresses a reset by multiplying the reset parameter by 0, whereas a ControlSignal with a non-zero value multiples the reset parameter’s non-zero default value, resulting in a non-zero value that elicits a reset.

Class Reference

class psyneulink.core.components.mechanisms.processing.integratormechanism.IntegratorMechanism(default_variable=None, input_shapes=None, input_ports=None, function=None, reset_default=0, output_ports=None, params=None, name=None, prefs=None, **kwargs)

Subclass of ProcessingMechanism that integrates its input. See Mechanism for additional arguments and attributes.

Parameters

  • function (IntegratorFunction : default IntegratorFunction) – specifies the function used to integrate the input. Must take a single numeric value, or a list or np.array of values, and return one of the same form.

  • reset_default (number, list or np.ndarray : default 0) – specifies the default value used for the reset parameter.

reset

if non-zero, the IntegratorMechanism’s reset method is called, which resets the value of the IntegratorMechanism to its initial value (see Resetting the IntegratorMechanism for additional details).

Type

int, float or 1d array of length 1 : default 0

_handle_default_variable(default_variable=None, input_shapes=None, input_ports=None, function=None, params=None)

If any parameters with len>1 have been specified for the Mechanism’s function, and Mechanism’s default_variable has not been specified, reshape Mechanism’s variable to match function’s, but make sure function’s has the same outer dimensionality as the Mechanism’s

_execute(variable=None, context=None, runtime_params=None, **kwargs)

Override to check for call to reset by ControlSignal

exception psyneulink.core.components.mechanisms.processing.integratormechanism.IntegratorMechanismError(message, component=None)