Package com.anji.nn

com.anji.nn contains classes implementing a simple artificial neural net framework.

See:
          Description

Interface Summary
ActivationFunction Abstracts activation function for neurons.
Connection Abstract connection to carry data into neurons.
 

Class Summary
ActivationFunctionFactory Factory to hide implementation details of creating activation functions and ensure there is always only one of each.
ActivationFunctionType Enumerated type representing flavors of neurons: input, output, hidden.
AnjiNet Aggregates all pieces for a full neural network.
BiasConnection Connection used for bias node, clamped at 1.
CacheNeuronConnection A connection between neurons that caches its value from previous timestep.
ClampedLinearActivationFunction  
EvSailSigmoidActivationFunction Modified classic sigmoid.
InverseAbsActivationFunction Inverse absolute value.
LinearActivationFunction Linear activation function.
Neuron Neuron component of an artificial neural network.
NeuronConnection Connection between neurons.
Pattern Converts double array stimuli to Connection objects.
RandomConnection  
RecurrencyPolicy Enumerated type representing policies for how to handle recurrency.
SigmoidActivationFunction Modified classic sigmoid.
SignedClampedLinearActivationFunction  
SignedStepActivationFunction Step activation function.
StepActivationFunction Step activation function.
StepHourglassConnection StepHourglassConnection Connection returning a value between 0 and 1, where 1 indicates the fully allotted amount of time remaining and 0 indicates no time remainnig.
TanhActivationFunction Hyperbolic tangent.
TanhCubicActivationFunction Hyperbolic tangent modified to have a "well" around 0.
 

Package com.anji.nn Description

com.anji.nn contains classes implementing a simple artificial neural net framework.