|
Neural Network Software | |||||||||
| PREV NEXT | FRAMES NO FRAMES | |||||||||
See:
Description
| Packages | |
| com.practicalstudies.neural | Neural network class library. |
| com.practicalstudies.neural.test | Test classes for the neural network class library. |
A class library for constructing artificial neural networks.
Networks are constructed as sets of Nodes connected by
WeightedConnections. In practice, the Nodes will be represented
by Neurons, suitably subclassed, using appropriate
Adder and ActivationFunction objects to calculate
Neuron output, and an appropriate WeightAdjustmentFunction
for network learning.
For backpropagation training, use BackpropagationNode, with
LinearAdder and SigmoidalActivationFunction
to calculate Neuron output, BackpropagationErrorFunction to
determine errors in Neuron output, and BackpropagationWeightAdjustment
to train the network. On reflection, this breakdown seems overly complex, and I am looking to rationalize it.
See BackpropagationTest as an example of how to use the indicated classes.
|
PracticalStudies.com | |||||||||
| PREV NEXT | FRAMES NO FRAMES | |||||||||