Minimizing the average cost of paging and registration: A timer-based method

Methods of balancing call registration and paging are developed in this paper. Given that the probability distribution on the user location as a function of time is either known or can be calculated, previous work shows the existence of lower bounds on the average cost of paging. Here these bounds are used in conjunction with a Poisson incoming-call arrival model to formulate the paging/registration optimization problem in terms oftimeout parameters, _m ; the maximum amount of time to wait before registering given the last known location wasm. Timer-based methods, as opposed to location-based methods, do not require the user to record and process location information during the time between location updates. This feature might be desirable for minimizing mobile transceiver use during idle periods. We then consider uniform motion processes where a spatial translation of starting location produces an identical spatial translation of the associated time-varying probability distribution. This leads to a universal timeout parameter which may be readily calculated. We study and the minimum cost of paging/registration for a simple model of user motion and compare our results to an earlier method of location-based paging/registration cost minimization.     

Twisting and annealing optical fiber for current sensors

We demonstrate that twisting a fiber a few turns per meter before it is annealed largely eliminates the residual linear birefringence. This dramatically improves the yield of annealed coils used for current sensing and makes it possible to use fibers that previously had large residual linear birefringence. Twisting the fiber is effective because the residual birefringence, associated with core ellipticity, is reduced to near zero by twisting. A theoretical model of the twisted and annealed fiber current sensor is compared to experimental data. We also show good temperature stability for a sensor made with this new technique     

Vector quantization by deterministic annealing

A deterministic annealing approach is suggested to search for the optimal vector quantizer given a set of training data. The problem is reformulated within a probabilistic framework. No prior knowledge is assumed on the source density, and the principle of maximum entropy is used to obtain the association probabilities at a given average distortion. The corresponding Lagrange multiplier is inversely related to the `temperature' and is used to control the annealing process. In this process, as the temperature is lowered, the system undergoes a sequence of phase transitions when existing clusters split naturally, without use of heuristics. The resulting codebook is independent of the codebook used to initialize the iterations