Loading...
Please wait, while we are loading the content...
Similar Documents
Measurement Design Framework for Network Tomography Using Fisher Information
| Content Provider | Semantic Scholar |
|---|---|
| Author | Liu, Chang He, Ting Swami, Ananthram Towsley, Don Salonidis, Theodoros Leung, Kin K. |
| Copyright Year | 2013 |
| Abstract | In this paper, we study the design of measurements to infer link parameters by measuring end-to-end performance along selected paths. While existing work focuses on the inference problem under given path measurements, the accuracy of inference fundamentally hinges on the set of available measurements. To this end, we propose a generic framework to design measurements for link parameter tomography to allow for most accurate inference of link parameters. Given a link model and a set of path measurements, it is known from estimation theory that the Fisher information matrix (FIM) characterizes a lower bound on the error of the optimal unbiased estimator, asymptotically achievable by the maximum likelihood estimator (MLE). We therefore formulate our measurement design problem to minimize a function of the FIM that represents the total estimation error. We apply this framework to inferring link loss rates from path losses. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://people.cs.umass.edu/~cliu/pdf/LiuEtal13AFMshort.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
