GANNet: A genetic algorithm for optimizing topology and weights in neural network design

White, David; Ligomenides, Panos A.

doi:10.1007/3-540-56798-4_167

Cited by 48 publications

(46 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The weight mutation is applied first given in [17], [11], [12]. Based on the concept of evolutionary algorithm, topological mutations are applied after evolution of weights.…”

Section: Topology Mutationmentioning

confidence: 99%

Simultaneous Evolution of Architecture and Connection Weights in Artificial Neural Network

Sagar¹,

Chalam²

2012

IJCA

View full text Add to dashboard Cite

The important issue for Designing architecture isthe evolution of Artificial Neural Network (ANN). There is no systematic method to design a near-optimal architecture for a given application or task. The pattern classification methods are used to design the neural network architectures and efforts towards the automatic design of network topologies, constructive and destructive algorithms can be used. In the proposed work the optimization of architectures and connection weights uses the evolutionary process. A singlepoint crossover is applied with selective schemas on the network space and evolution is introduced in the mutation stage so that an optimized ANNs are achieved.

show abstract

“…The weight mutation is applied first given in [17], [11], [12]. Based on the concept of evolutionary algorithm, topological mutations are applied after evolution of weights.…”

Section: Topology Mutationmentioning

confidence: 99%

Simultaneous Evolution of Architecture and Connection Weights in Artificial Neural Network

Sagar¹,

Chalam²

2012

IJCA

View full text Add to dashboard Cite

show abstract

“…A binary encoding with variable number of weighed bits was used which allowed the shortening of the chromosome at the small weigh value. The authors of [10] [11] proposed an innovative method which would create networks that were not multilayer perceptrons as generally acknowledged. The information about the connections between a given neuron and other neurons was encoded in the chromosome.…”

Section: Earlier Attemptsmentioning

confidence: 99%

Application of genetically evolved neural networks to dynamic terrain generation

Chomatek¹,

Rudnicki²

2011

Bulletin of the Polish Academy of Sciences: Technical Sciences

View full text Add to dashboard Cite

Abstract. Real time terrain generation is a vital part in the development of realistic computer simulations and games. Dynamic terrain generation influences the realism of simulation, because its participants have to adapt to the current environment conditions. Dynamically generated primary terrain is transformed in order to reflect natural phenomena, such as thermal and water erosion, avalanches or glaciers. In this article a possibility of primary terrain transformation with application of artificial neural networks is shown. The networks are trained by evolutionary algorithms to solve a problem of a water erosion phenomenon. Obtained results show that application of such neural networks to this problem can significantly reduce the processing time needed to perform the process of modeling the natural phenomena.

show abstract

“…The transfer function is often assumed to be same for all the nodes in the same layer. White et al [27] introduced a simple approach to evolve both architectures and connection weights. The evolution was used to decide the optimal mixture between two transfer functions (sigmoid transfer function and Gaussian transfer function).…”

Section: Optimization Of the Drnn Structurementioning

confidence: 99%

“…The transfer function converges as a log sigmoid. For comparison, the GANNet algorithm presented in [27] was implemented to evolve the architecture of the DRNN. Because the GA is sensitive to the length of the chromosome string, the shorter this string can be made, the better, the GANNet was modified to optimize the architecture of the DRNN; then the DBP was used to obtain the weights.…”

Section: Evolution Processmentioning

confidence: 99%

Development of a new EDRNN procedure in control of human arm trajectories

2008

View full text Add to dashboard Cite

GANNet: A genetic algorithm for optimizing topology and weights in neural network design

Cited by 48 publications

References 5 publications

Simultaneous Evolution of Architecture and Connection Weights in Artificial Neural Network

Simultaneous Evolution of Architecture and Connection Weights in Artificial Neural Network

Application of genetically evolved neural networks to dynamic terrain generation

Development of a new EDRNN procedure in control of human arm trajectories

Contact Info

Product

Resources

About