Bandwidth variability prediction with rolling interval least squares (RILS)

Abstract

1 : Real Time prediction of end-to-end bandwidth is crucial for network quality of service. It remains a challenge for providing prediction in good quality. This is mainly because of uncertainties involved in network communication. Very recently a new algorithm called Rolling Interval Least Squares (RILS) [3] has been developed which significantly improved the variability forecast of the stock market. In this paper we extend the interval computing approach and applied RILS to innovatively predict bandwidth based on a wide range of rates. Experimental results indicate that RILS improved the quality of bandwidth variability prediction by about 24% when compared to current methods.

Managing Uncertainty in Crowdsourcing with Interval-Valued Labels

“…It has been further verified and validated from the perspective of information theory that the significant improvement comes indeed from interval-valued data [10]. Other successes include but are not limited to [3,6,[11][12][13][14]17], and many more. Past successes motivate us to apply interval-valued labels in crowdsourcing.…”