Package network.aika.neuron.conjunctive
Class ConjunctiveNeuron<S extends ConjunctiveSynapse,A extends ConjunctiveActivation>
java.lang.Object
network.aika.neuron.Neuron<S,A>
network.aika.neuron.conjunctive.ConjunctiveNeuron<S,A>
- All Implemented Interfaces:
Writable
- Direct Known Subclasses:
BindingNeuron
,PatternNeuron
public abstract class ConjunctiveNeuron<S extends ConjunctiveSynapse,A extends ConjunctiveActivation>
extends Neuron<S,A>
- Author:
- Lukas Molzberger
-
Field Summary
Fields inherited from class network.aika.neuron.Neuron
allowTraining, bias, frequency, inputSynapses, isNetworkInput, lock, outputSynapses, sampleSpace
-
Constructor Summary
-
Method Summary
Modifier and TypeMethodDescriptionvoid
protected void
void
If the complete bias exceeds the threshold of 0 by itself, the neuron would become constantly active.void
Methods inherited from class network.aika.neuron.Neuron
addInputSynapse, addOutputSynapse, applyMovingAverage, biasUpdate, checkTemplateInductionThreshold, containsInputSynapse, containsOutputSynapse, count, createActivation, getActivations, getBias, getCandidateGradient, getCustomData, getFrequency, getFrequency, getId, getInputSynapse, getInputSynapses, getLabel, getLock, getModel, getOutputSynapse, getOutputSynapses, getProbability, getProvider, getRetrievalCount, getSampleSpace, getSurprisal, getTargetSynapses, getTemplate, getTemplateGroup, getTemplateInfo, initFromTemplate, instantiateTemplate, isAllowTraining, isModified, isNetworkInput, isTemplate, reactivate, read, readFields, register, removeInputSynapse, removeOutputSynapse, resetModified, setAllowTraining, setCustomData, setFrequency, setLabel, setModified, setNetworkInput, setProvider, suspend, templateNeuronMatches, toKeyString, toString, write
-
Constructor Details
-
ConjunctiveNeuron
public ConjunctiveNeuron() -
ConjunctiveNeuron
-
ConjunctiveNeuron
-
-
Method Details
-
initFromTemplate
-
limitBias
public void limitBias()If the complete bias exceeds the threshold of 0 by itself, the neuron would become constantly active. The training should account for that and reduce the bias back to a level, where the neuron can be blocked again by its input synapses.- Overrides:
limitBias
in classNeuron<S extends ConjunctiveSynapse,
A extends ConjunctiveActivation>
-
addInactiveLinks
- Specified by:
addInactiveLinks
in classNeuron<S extends ConjunctiveSynapse,
A extends ConjunctiveActivation>
-
getActivationFunction
- Specified by:
getActivationFunction
in classNeuron<S extends ConjunctiveSynapse,
A extends ConjunctiveActivation>
-
updateAllowPropagate
public void updateAllowPropagate()- Overrides:
updateAllowPropagate
in classNeuron<S extends ConjunctiveSynapse,
A extends ConjunctiveActivation>
-