A Detection Scheme for Dependent Noise Processes

In this paper the problem of memoryless discrete time detection of a constant signal in additive φ-mixing noise is considered. The fidelity criterion is taken to be the asymptotic relative efficiency. The optimal detector nonlinearity is approximated by a polynomial, and the statistical knowledge required for the design of the detector is the family of joint moments of the noise, resulting in a very simple design procedure for the detector. It is shown that the performance of this detector can be made arbitrarily close to that of the optimal memoryless detector for a large class of noises.     

On metrics matched to the discrete memoryless channel

A sequence of metrics {D_N} is said to be additive and matched to a discrete memoryless channel (DMC) if D_N is the sum on its coordinates of N single letter metrics and if the maximum likelihood decoder for sequences of length N is a minimum D_N-distance decoder. Necessary and sufficient conditions on the transition probabilities of a DMC for the existence of a sequence of additive metrics matched to it are given. In the case of the binary channel these are shown to be equivalent to the channel being symmetric. Explicit transition probabilities are given for a large class of ternary DMCs with an associated sequence of additive matched metrics. The problem solved here may be considered a generalization of the problem of finding the DMCs matched to the Lee metric solved by Chiang and Wolf in 1971 (2).     

A control problem with structural choices

The case is considered in which, during the operation of an optimal control system, the optimizer, in addition to applying his usual control, may switch structures. Necessary and sufficient conditions are derived and emphasis is placed on the special characteristics of this problem. Continuous and discrete time set-ups are considered and the separation principle is shown not to hold for the linear quadratic case in the presence of noise.     

Dimensions of R&D location in the United States

R&D activities in the United States, as in other advanced economies, are geographically concentrated in certain types of locations. This study presents data on the location of four dimensions of R&D in the U.S.: industrial R&D laboratories, scientists and engineers engaged in R&D, scientists and engineers employed by the federal government, and research universities. Industrial R&D is much more concentrated in large urban areas than the other dimensions, and appears to locate more in response to the location of manufacturing activity than to the location of research universities and federal research facilities. The location of R&D employment, which includes government university, and industrial employees, is associated with facilities for all three types of R&D. Because of these factors, R&D in the U.S. is found on a significant per capita basis in 44 of 177 urban areas, most of them in the northeastern portion of the country. When two dimensions, industrial R&D laboratories and R&D employees, are combined as a measure of R&D concentration, the locational pattern is less clustered regionally. Ten urban areas in all regions of the U.S. are identified as important complexes of R&D. Since the location of R&D is a major indicator of comparative advantage for technological activities and the economic potential of urban regions, only a few areas of the U.S. are likely to remain important in the generation of innovations.     

Water-clocks and time measurement in classical antiquity

The water-clock was the first mechanical device for time measurement to be used by the Greeks. Previously they had relied on sundials, or observation of the phases of the moon or the position of the sun in the zodiac to locate points of time within the day, month or year; and by the mid-fifth century BC they had made some progress with the difficult calendar problems due to the fact that the solar day, the lunar month, and the solar year are not commensurable with each other in whole numbers.     

The graph-theoretic field model—I. Modelling and formulations

The Graph-Theoretical Field Model provides a unifying approach for developing numerical models of field and continuum problems. The methodology examines the field problem from the first stages of conceptualization without recourse to the governing differential equations of the field problem; this is accomplished by deriving discrete statements of the physical laws which govern the field behaviour. There are generally three laws, and these are modelled by the “cutset equations”, the “circuit equations”, and the “terminal equations”. In order to establish these three sets of equations it is expedient first to spatially discretize the field in a manner similar to the finite difference method and then to associate a linear graph (denoted as the field graph) with the spatial discretization. The concept of “through” and “across” variables, which underlies the cutset and circuit equations respectively, enables one to define the graph in an unambiguous manner such that each “edge” of the graph identifies a pair of complementary variables. From a knowledge of the constitutive properties and the boundary conditions of the field it is possible to associate terminal equations with sets of edges. Since the resulting sets of equations represent the field equations, these equations provide the basis for a complete (but approximate) solution to the field or continuum problem. In fact, this system approach uses a two part model: one for the components and another for the interconnection pattern of the components which renders the formulation procedures totally independent of the solution procedure.This paper presents the theoretical basis of the model and several graph-theoretic formulations for steady-state problems. Examples from heat conduction and small- deformation elasticity are included.