Fast Information Detection in Big Data using Neural Networks and Matrix Decomposition

Authors: ,

Abstract: In previous work, fast neural networks (FNNs) for information extraction was presented. The operation of these networks relies on performing cross-correlation between the input patterns and the weights of processing elements in the frequency domain. In this paper, a new strategy to accelerate this approach is introduced. Such strategy applies the concept of divide and conquer to reduce the number of calculation steps required by FNNs. The big data matrix is decomposed into smaller sub-matrices. Each generated sub-matrix is processed by using a single FNN implemented in the frequency domain. As a result, the speed up ratio is increased with the size of the input big data matrix. This is in contrast to using only FNNs. Simulation results show that the proposed approach for information detection in big data is faster than the conventional neural networks and FNNs. Moreover, experimental results for big data matrices with different sizes show good performance.

WSEAS Transactions on Systems, ISSN / E-ISSN: 1109-2777 / 2224-2678, Volume 15, 2016, Art. #6

PDF
XML
Cite
×
Citation Tools
WSEAS AMA
Hazem M. El-Bakry, Nikos Mastorakis, "Fast Information Detection in Big Data using Neural Networks and Matrix Decomposition," WSEAS Transactions on Systems, vol. 15, pp. 49-58, 2016, DOI:
Hazem M. El-Bakry, Nikos Mastorakis. Fast Information Detection in Big Data using Neural Networks and Matrix Decomposition. WSEAS Transactions on Systems. 2016;15:49-58.
Copy to Clipboard
Citation copied to Clipboard

Certification

---