Reference and Bibtex for Evolving neural networks through augmenting topologies by Stanley, Kenneth O. et. al.

@article{638554,
 author = {Kenneth O. Stanley and Risto Miikkulainen},
 title = {Evolving neural networks through augmenting topologies},
 journal = {Evol. Comput.},
 volume = {10},
 number = {2},
 year = {2002},
 issn = {1063-6560},
 pages = {99--127},
 doi = {http://dx.doi.org/10.1162/106365602320169811},
 publisher = {MIT Press},
 address = {Cambridge, MA, USA},
 }

Language:
  en
Type:
  article
Title:
  Evolving neural networks through augmenting topologies
Author:
  Stanley, Kenneth O. et. al.
Topic:

Rating:
  4
Comment:

uses direct coding, since, cite: "Although CE [Gruau s (1993) Cellular Encoding (CE)] demonstrates that it is possible to evolve developmental systems, we chose direct encoding for NEAT because, as Braun andWeisbrod (1993) argue, indirect encoding requires  more detailed knowledge of genetic and neural mechanisms.  In other words, because indirect encodings do not map directly to their phenotypes, they can bias the search in unpredictable ways. To make good use of indirect encodings, we need to first understand them well enough to make sure that they do not focus the search on some suboptimal class of topologies. Further, experimental results suggest that CE is not necessarily more efficient than direct encoding methods (Section 4.3.3)."

describes "historical marking" to allow for save crossover (but needs global counter) - ligning up genes with equal number; genes which are only present in one parent are either ignored (less fit) or completely used (more fit parent)

"explicit fitness sharing" of similar genotypes to facilitate speciation (each on a different "peak")

mutation always adds (though genes might become disabled); either synapse or neuron (split connection)

pole-balancing as benchmark, despite of huge similarities to ENS^3 no reference to Pasemann results (but source code available!)

Acquired:
2005-06-28

Reference:

Bibtex:

Other information: