arXiv subjects

Kerry Fendick

Publications and source records attributed to Kerry Fendick.

6 recordsLinked to original sources

arXiv2019

Improving Conversation Quality for VoIP Through Block Erasure Coding

The conversational quality of voice over IP (VoIP) depends on packet-loss rates, burstiness of packet loss, and delays (or latencies). The benefits for conversational voice quality of erasure coding attributable to its reduction in packet loss rates are widely appreciated. When block erasure coding is used, our analysis shows how those benefits are reduced or even eliminated by increases in delays and in a measure of burstiness of packet loss. We nevertheless show that the net effect of those three factors is still positive over a wide range of network loss rates provided that block sizes are sufficiently small and the sizes of decoding buffers have been optimized for real-time media. To perform this analysis, we develop a new analytical model describing the effects of block erasure coding on end-to-end network performance.

cs.IT↗

arXiv2019

The Effect of Erasure Coding on the Burstiness of Packet Loss

The perceived quality of real-time media delivered over IP networks depends on both the rate and burstiness of packet loss. In this paper, we develop a new mathematical model for the residual burstiness of loss under erasure coding. We derive the expected number of consecutive losses in a burst as a function of erasure coding parameters and the network loss probability assuming a Bernoulli model for network losses.

cs.IT↗

arXiv2016

Gaussian Processes for Local Polynomial Forecasting of Time Series

Non-stationary time series with non-linear trends are frequently encountered in applications. We consider here the feasibility of accurately forecasting the signals of multiple such time series considering jointly when the number of historic samples is inadequate for accurately forecasting the signal of each considered in isolation. We develop a new forecasting methodology based on Gaussian process regression that is successful in doing so in examples for which the method of generalized least-squares is not. The new method employs a form of deep machine learning.

stat.ME↗

arXiv2016

Covariance Kernels of Gaussian Markov Processes

The solution to a multivariate linear Stochastic Differential Equation (SDE) with constant initial state is well known to be a Gaussian Markov process, but its covariance kernel involves the solution to an integral equation in the general case. We show that the covariance kernel has a simpler semi-parametric form for families of such solutions representing increments of a common process. We also show that a covariance kernel of a particular parametric form is necessary and sufficient for a solution to possess stationary increments and for a Gaussian process, in considerable generality, to have stationary increments and the Markov property. For a discretely sampled Gaussian process with such a parametric kernel, we derive closed-form expressions for unique maximum likelihood estimators of the parameter matrices that are unbiased, jointly sufficient, and easily computed regardless of the dimension. Using those estimators, we also derive closed-form expressions for posterior moments useful for forecasting.

math.PR↗

arXiv2014

Queues with Censored Demand and Autoregressive Net Input

We develop methods for simulating a queue with an autoregressive net-input process and for recovering characteristics of such a net-input process from samples of queue lengths. We apply these methods to the problem of estimating the censored (unsatisfied) demand for the queue's content and show how to model a queue for which the censoring of demand is graduated in a neighborhood of the queue's zero lower bound. As an example, we estimate the monthly unsatisfied demand for U.S. nonfarm jobs based on samples of job openings through a period including the last two recessions.

math.PR↗

arXiv2012

Gaussian Fluid Queue with Autocorrelated Input

This paper develops a generalization of Brownian motion with stationary, autocorrelated increments as a tractable model for problems in business and finance. We show that any real continuous Gaussian Markov process with stationary increments and smooth covariance function is characterized by three parameters quantifying drift, volatility, and autocorrelations. We model a queue as a functional of a process defined by those characteristics and derive its transient distribution conditional on its history.

math.PR↗