Nec Research Institute, Inc.

Nec Usa, Inc.

Benjamin Melamed

Predrag Jelenkovic

David Jagerman

TES (Transform-Expand-Sample) is a versatile class of stationary stochastic processes which can model arbitrary marginals, a wide variety of autocorrelation functions, and a broad range of sample path behaviors. An algorithmic method replaces the previously relied upon heuristic search, thereby automating TES modeling for simulation analysis. The algorithm is solved in a nonlinear programming setting, which takes advantage of fast and accurate computation of TES autocorrelation functions and their partial derivatives to implement a steepest-descent technique, preferably based on Zoutendijk's Feasible Direction Method. The method has particular application to data compression and specifically compressed video.

Algorithmic modeling of TES processes for simulation analysis

To provide a time series having correlated interarrival times for use in traffic and queueing studies, an independent identically distributed random number stream is first transformed to a sequence of correlated uniform variates by a special modulo-1 autoregressive operation and then further transformed into a sequence of correlated variates with a prescribed marginal distribution from which there is developed the desired time studies.

Growing community of inventors

Warren, NJ, United States of America

Benjamin Melamed