Anzeige
Erschienen in:
2016 | OriginalPaper | Buchkapitel
Short Latency Bias in Latency Matrix Completion
verfasst von : Cong Wang, Min LI, Yan Yang
Erschienen in: Frontier Computing
Verlag: Springer Singapore
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
For latency-sensitive applications, a key issue is how to estimate the latencies between any couple of nodes. Latency Matrix Completion method provides a simple but efficient way to estimate the latencies instead of measure them directly. In this paper, we make comparative studies on several Internet latency data sets, and report an easy overlooked shortcoming exists in Latency Matrix Completion. For short latencies, their relative estimation errors are much higher than those of long latencies. In this paper, we propose a brief model to analyze why this bias exists. We believe that the loss function which used in the optimizing process is a possible reason for this phenomenon. How to remove the short latency bias should cause our consideration in the future.