Structural and Behavioral Evolution of Recurrent Networks

Gregory M. Saunders, Peter J. Angeline, Jordan B. Pollack

NeurIPS 1993 pp. 88-95

/neurips/1993/saunders1993neurips-structural/

Abstract

This paper introduces GNARL, an evolutionary program which induces recurrent neural networks that are structurally unconstrained. In contrast to constructive and destructive algorithms, GNARL employs a popula(cid:173) tion of networks and uses a fitness function's unsupervised feedback to guide search through network space. Annealing is used in generating both gaussian weight changes and structural modifications. Applying GNARL to a complex search and collection task demonstrates that the system is capable of inducing networks with complex internal dynamics.

PDF NeurIPS Semantic Scholar

Cite

Text

Saunders et al. "Structural and Behavioral Evolution of Recurrent Networks." Neural Information Processing Systems, 1993.

Markdown

[Saunders et al. "Structural and Behavioral Evolution of Recurrent Networks." Neural Information Processing Systems, 1993.](https://mlanthology.org/neurips/1993/saunders1993neurips-structural/)

BibTeX

@inproceedings{saunders1993neurips-structural,
  title     = {{Structural and Behavioral Evolution of Recurrent Networks}},
  author    = {Saunders, Gregory M. and Angeline, Peter J. and Pollack, Jordan B.},
  booktitle = {Neural Information Processing Systems},
  year      = {1993},
  pages     = {88-95},
  url       = {https://mlanthology.org/neurips/1993/saunders1993neurips-structural/}
}