Package network.aika.steps.activation
Class Propagate
The job of the linking phase is to propagate information through the network by creating the required activations and links.
Each activation and each link have an corresponding neuron or synapse respectively. Depending on the data set in the
document, a neuron might have several activations associated with it. During propagation an input activation
causes the creating of a link in one or more output synapses and the creation of an output activation. Initially the value
of the input activation and the weight of the synapse might not suffice to activate the output activation. But that might
change later on as more input links are added to the activation. New input links are added by the closeCycle method. This
method is called by the visitor which follows the links in the activation network to check that both input and output
activation of a new link refer to the same object in the input data set.
- Author:
- Lukas Molzberger
-
Field Summary
Fields inherited from interface network.aika.steps.QueueKey
COMPARATOR, ELEMENT_COMPARATOR, THOUGHT_COMPARATOR
-
Method Summary
Methods inherited from class network.aika.steps.Step
add, checkIfQueued, copy, getElement, getFired, getStepName, getTimestamp, setTimestamp
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait
Methods inherited from interface network.aika.steps.QueueKey
getCheckIfQueuedTimestamp
-
Method Details
-
add
public static void add(Activation act, boolean template, String linkingType, Predicate<Synapse> filter) -
getPhase
- Specified by:
getPhase
in interfaceQueueKey
- Specified by:
getPhase
in classStep<Activation>
-
getLinkingOrder
- Specified by:
getLinkingOrder
in interfaceQueueKey
- Overrides:
getLinkingOrder
in classStep<Activation>
-
process
public void process()- Specified by:
process
in classStep<Activation>
-
propagate
-
toString
- Overrides:
toString
in classStep<Activation>
-