Parallelization of a series of extreme learning machine algorithms based on spark | IEEE Conference Publication | IEEE Xplore

Parallelization of a series of extreme learning machine algorithms based on spark


Abstract:

With the development of the Internet, traditional big data computing platform gradually lose its competitive advantages as a result of high latency. On the contrary, a fa...Show More

Abstract:

With the development of the Internet, traditional big data computing platform gradually lose its competitive advantages as a result of high latency. On the contrary, a fast, easy to use and generic big data computing frame called Spark draws more and more attentions. At the same time the integrated solution which is based on RDD (Resilient Distributed Datasets) and is offered by Spark makes the applications of Spark in actual projects broader and broader. Non-iterative ELM (Extreme Learning Machine) algorithm which generates hidden layer weights randomly determines the output layer weights by analyzing. Using this method to reduce learning time as more as possible can bring much convenience to many time-sensitive applications. In this article we put forward a kind of Feedforward Neural Network Parallel Algorithm which is based on Spark platform, establish VMware vSphere platform, experiments on vSphere VMware experiment platform. Our experiment results show that this algorithm can increase the analysis speed of ELM algorithm.
Date of Conference: 26-29 June 2016
Date Added to IEEE Xplore: 25 August 2016
ISBN Information:
Conference Location: Okayama, Japan

References

References is not available for this document.