A back-propagation (BP) neural network can solve complicated random<br />nonlinear mapping problems; therefore, it can be applied to a wide range<br />of problems. However, as the sample size increases, the time required to<br />train BP neural networks becomes lengthy. Moreover, the classification<br />accuracy decreases as well. To improve the classification accuracy and<br />runtime efficiency of the BP neural network algorithm, we proposed a<br />parallel design and realization method for a particle swarm optimization<br />(PSO)-optimized BP neural network based on MapReduce on the Hadoop<br />platform using both the PSO algorithm and a parallel design. The PSO<br />algorithm was used to optimize the BP neural network's initial weights<br />and thresholds and improve the accuracy of the classification algorithm.<br />The MapReduce parallel programming model was utilized to achieve<br />parallel processing of the BP algorithm, thereby solving the problems of<br />hardware and communication overhead when the BP neural network addresses<br />big data. Datasets on 5 different scales were constructed using the<br />scene image library from the SUN Database. The classification accuracy<br />of the parallel PSO-BP neural network algorithm is approximately 92%,<br />and the system efficiency is approximately 0.85, which presents obvious<br />advantages when processing big data. The algorithm proposed in this<br />study demonstrated both higher classification accuracy and improved time<br />efficiency, which represents a significant improvement obtained from<br />applying parallel processing to an intelligent algorithm on big data.
Cao, J., Cui, H., Shi, H., & Jiao, L. (2016). Big data: A parallel particle swarm optimization-back-propagation neural network algorithm based on MapReduce. PLoS ONE, 11(6). https://doi.org/10.1371/journal.pone.0157551