- All Implemented Interfaces:
- java.io.Serializable, Speciation
public class OriginalNEATSpeciation
extends ThresholdSpeciation
The original NEAT Speciation Strategy. This is currently the only speciation
strategy implemented by Encog. There are other speciation strategies that
have been proposed (and implemented) for NEAT. One example is k-means.
NEAT starts up by creating an initial population of genomes with randomly
generated connections between input and output nodes. Not every input neuron
is necessarily connected, this allows NEAT to determine which input neurons
to use. Once the population has been generated it is speciated by iterating
over this population of genomes. The first genome is placed in its own
species.
The second genome is then compared to the first genome. If the compatibility
is below the threshold then the genome is placed into the same species as the
first. If not, the second genome founds a new species of its own. The
remaining genomes follow this same process.
-----------------------------------------------------------------------------
http://www.cs.ucf.edu/~kstanley/ Encog's NEAT implementation was drawn from
the following three Journal Articles. For more complete BibTeX sources, see
NEATNetwork.java.
Evolving Neural Networks Through Augmenting Topologies
Generating Large-Scale Neural Networks Through Discovering Geometric
Regularities
Automatic feature selection in neuroevolution
- See Also:
- Serialized Form