Genetic Algorithm Trained Counter-Propagation Neural Net in Structural Optimization

Iranmanesh, Amir; Fahimi, Mansour

doi:10.4203/ccp.74.36

Cited by 1 publication

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Basically, from the perspective of the current work, different alternatives of SOMs as well as PSO could be used for mapping as well as classification task. Some of those alternatives are ISOMAP [59], manifold alignment [60], supervised kohonen network (SKN) [62], counter propagation artificial neural network [61], support vector machine (SVM) [63] and, principle component analysis (PCA) [64]. Such algorithm can be employed to estimate the SOMs, which proved that the SOMs outperforms all the mapping based algorithms where PSO shows better performance with the increasing number of complex as well as complicated datasets.…”

Section: Introductionmentioning

confidence: 99%

“…Alongside, SOMs lead to higher sensitivity for clustering and projection purpose. Also, manipulation of input data under SOMs is unsupervised and automatic, while counter propagation artificial neural network (CP-ANN) [61], and supervised kohonen networks (SKN) [62] are based on supervised learning. This is the key advantage for SOMs because it can automatically adjust its parameters for any kinds of input data and no further supervision is necessary.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Self-organizing mapping based swarm intelligence for secondary and tertiary proteins classification

Kamal

Sarowar

Dey³

et al. 2017

Int. J. Mach. Learn. & Cyber.

View full text Add to dashboard Cite

Proteins have a significant role in animals and human health. Interactions among proteins are complex and large. Proteins separations are challenging process in molecular biology. Computational tools help to simulate the analysis in order to reduce the training data into small testing data. Large proteins have been mapped using self-organizing maps (SOMs). Neural network based SOMs has a significant role in reducing the irregular shapes of proteins interactions. Iterative checking enables the organizations of all proteins. In next stage, particle swarm intelligence is applied to classify the proteins' families. In the current work, secondary (Two dimensional) and tertiary proteins (Three dimensional) proteins have been grouped. Two dimensional proteins contain fewer hydrocarbons than three dimensional proteins. For faster analysis, the angles of the proteins are taken into account. The SOMs is compared with Bounding Box approach. In final, the experimental evolutions show that swarm intelligence achieved faster processing through enabling less memory consumptions and time. Since PSO combines proteins datasets in fuzzy values, the compactness or integration of similar proteins are strong. On the other hand, Bounding Box uses the Crisp value. Therefore, it needs more space to organize the whole data. Without SOMs, swarm intelligence also results are poor due to the excessive time consuming and required storage area. Moreover, for almost all classification and clustering tools, it is observed that the overall classification task becomes slow, time consuming, space consuming and also less sensitive because of noises, irrelevant data in input datasets. Thus, the proposed SOM based PSO approach achieved less time consuming with efficient classification into secondary and tertiary proteins.

show abstract