Evolutionary deep neural networks

Research output: Chapter in Book/Report/Conference proceedingChapter

2 Citations (Scopus)

Abstract

This chapter first employs a kinematic model of hand to create two datasets for static hand postures. Neural Networks with different learning algorithms are then applied to the datasets for classification. The chapter also considers the comparison and analysis of different evolutionary algorithms for classifying datasets as well. Another contribution is finding the best set of features for the dataset using evolutionary algorithms. The results show that due to the large number of samples and features, Back Propagation is not effective due to the problem of local optima stagnation. However, Evolutionary Algorithms are able to efficiently classify the dataset with a very high accuracy and convergence speed. It was also observed that feature selection is important and evolutionary algorithms are able to find the optimal set of features for this problem.

Original languageEnglish
Title of host publicationStudies in Computational Intelligence
PublisherSpringer Verlag
Pages141-156
Number of pages16
DOIs
Publication statusPublished - 1 Jan 2019
Externally publishedYes

Publication series

NameStudies in Computational Intelligence
Volume780
ISSN (Print)1860-949X

    Fingerprint

Cite this

Mirjalili, S. (2019). Evolutionary deep neural networks. In Studies in Computational Intelligence (pp. 141-156). (Studies in Computational Intelligence; Vol. 780). Springer Verlag. https://doi.org/10.1007/978-3-319-93025-1_9