Search

In this paper, we present a new tree mining algorithm, DRYADEPARENT, based on the hooking principle first introduced in DRYADE. In the experiments, we demonstrate that the branching factor and depth of the frequent patterns to find are key factors of complexity for tree mining algorithms, even if often overlooked in previous work. We show that DRYADEPARENT outperforms the current fastest algorithm, CMTreeMiner, by orders of magnitude on data sets where the frequent tree patterns have a high branching factor. Index Terms--Data mining, mining methods and algorithms, mining tree structured data.

Progressively emphasizing samples that are difficult to classify correctly is the base for the recognized high performance of real Adaboost (RA) ensembles. The corresponding emphasis function can be written as a product of a factor that measures the quadratic error and a factor related to the proximity to the classification border; this fact opens the door to explore the potential advantages provided by using adjustable combined forms of these factors. In this paper, we introduce a principled procedure to select the combination parameter each time a new learner is added to the ensemble, just by maximizing the associated edge parameter, calling the resulting method the dynamically adapted weighted emphasis RA (DW-RA). A number of application examples illustrates the performance improvements obtained by DW-RA. Index Terms--Adaboost, boosting, convex combination, dynamic parameter selection, emphasis.

An innovative approach for the optimal matching of independently optimum sum and difference patterns through sub-arrayed monopulse linear arrays is presented. By exploiting the relationship between the independently optimal sum and difference excitations, the set of possible solutions is considerably reduced and the synthesis problem is recast as the search of the best solution in a noncomplete binary tree. Towards this end, a fast resolution algorithm that exploits the presence of elements more suitable to change subarray membership is presented. The results of a set of numerical experiments are reported in order to validate the proposed approach pointing out its effectiveness also in comparison with state-of-the-art optimal matching techniques. Index Terms--Linear arrays, monopulse antennas, sum and difference pattern synthesis, tree-searching algorithm.

Data association which obtains relationship between radar measurements and existing tracks plays one important role in radar multiple-target tracking (MTT) systems. A new approach to data association based on the competitive Hopfield neural network (CHNN) is investigated, where the matching between radar measurements and existing target tracks is used as a criterion to achieve a global consideration. Embedded within the CHNN is a competitive learning algorithm that resolves the dilemma of occasional irrational solutions in traditional Hopfield neural networks. Additionally, it is also shown that our proposed CHNN-based network is guaranteed to converge to a stable state in performing data association and the CHNN-based data association combined with an MTT system demonstrates target tracking capability. Computer simulation results indicate that this approach successfully solves the data association problems.

We consider the problem of waveform design for multiple input/multiple output (MIMO) radars, where the transmit waveforms are adjusted based on target and clutter statistics. A model for the radar returns which incorporates the transmit waveforms is developed. The target detection problem is formulated for that model. Optimal and suboptimal algorithms are derived for designing the transmit waveforms under different assumptions regarding the statistical information available to the detector. The performance of these algorithms is illustrated by computer simulation.

By using a linear antenna array, velocity synthetic aperture radar (VSAR) can detect, focus, and locate slowly moving targets well. However, it may mis-locate fast moving targets in the azimuth (cross-range) direction. In this correspondence, we propose a synthetic aperture radar (SAR) with a nonuniform linear antenna array and give a design of the antenna arrangement. It is shown that our proposed nonuniform linear antenna array SAR (NULA-SAR) can locate both slowly and fast moving targets correctly. An integrated NULA-SAR algorithm for moving target imaging is also presented, and it is verified by some simulations.

We consider radar waveform selection algorithms for tracking accelerating targets. Tracking is handled within an interacting multiple model (IMM) framework, which produces a number of potential cost functions for minimization. The waveform library is obtained by implementing a fractional Fourier transform on a base waveform, thereby rotating its ambiguity function around its peak. This has the effect of rotating the lower bound of measurement covariance in the range-Doppler plane. Simulation results consider a number of potential cost functions, and show robustness to cases in which the target modes are not precisely known.

It is of great importance to develop a robust and fast tracking algorithm in passive localization and tracking system because of its inherent disadvantages such as weak observability and large initial errors. In this correspondence, a new algorithm referred to as the iterated unscented Kalman filter (IUKF) is proposed based on the analysis and comparison of conventional nonlinear tracking problem. The algorithm is developed from UKF but it can obtain more accurate state and covariance estimation. Compared with the traditional approaches (e.g., extended Kalman filter (EKF) and UKF) used in passive localization, the proposed method has potential advantages in robustness, convergence speed, and tracking accuracy. The correctness as well as validity of the algorithm is demonstrated through numerical simulation and experiment results.

Among target tracking algorithms using Kalman filtering-like approaches, the standard assumptions are Gaussian process and measurement noise models. Based on these assumptions, the Kalman filter is widely used in single or multiple filter versions (e.g., in an interacting multiple model (IMM) estimator). The oversimplification resulting from the above assumptions can cause degradation in tracking performance. In this paper we explore the application of Kalman-Levy filter to handle maneuvering targets. This filter assumes a heavy-tailed noise distribution known as the Levy distribution. Due to the heavy-tailed nature of the assumed distribution, the Kalman-Levy filter is more effective in the presence of large errors that can occur, for example, due to the onset of acceleration or deceleration. However, for the same reason, the performance of the Kalman-Levy filter in the nonmaneuvering portion of track is worse than that of a Kalman filter. For this reason, an IMM with one Kalman and one Kalman-Levy module is developed here. Also, the superiority of the IMM with Kalman-Levy module over only Kalman-filter-based IMM for realistic maneuvers is shown by simulation results.

A star pattern identification algorithm, called modified grid algorithm, for attitude determination of spacecraft is addressed. The proposed algorithm is closely connected to a general pattern recognition technique. Each star is characterized by a well-defined pattern that can be determined by the surrounding stars. The so-called grid algorithm is one of the star identification algorithms based upon the pattern recognition approach. The proposed algorithm, is motivated by the conventional grid algorithm. To enhance the performance of the conventional grid algorithm, a modified method using polar grid, virtual grid, and multi-references is proposed. Simulation study is conducted for the demonstration of the new algorithm. The proposed modified grid approach turns out to make the grid algorithm more robust and reliable.

A common problem in multi-target tracking is to approximate a Gaussian mixture by one containing fewer components; similar problems can arise in integrated navigation. A common approach is successively to merge pairs of components, replacing the pair with a single Gaussian component whose moments up to second order match those of the merged pair. Salmund [1] and Williams [2, 3] have each proposed algorithms along these lines, but using different criteria for selecting the pair to be merged at each stage. The paper shows how under certain circumstances each of these pair-selection criteria can give rise to anomalous behaviour, and proposes that a key consideration should be the Kullback-Leibler (KL) discrimination of the reduced mixture with respect to the original mixture. Although computing this directly would normally be impractical, the paper shows how an easily computed upper bound can be used as a pair-selection criterion which avoids the anomalies of the earlier approaches. The behaviour of the three algorithms is compared using a high-dimensional example drawn from terrain-referenced navigation.

Space systems are characterized by a low-intensity process noise resulting from uncertain forces and moments. In many cases, their scalar measurement channels can be assumed to be independent, with one-dimensional internal dynamics. The nominal operation of these systems can be severely damaged by faults in the sensors. A natural method that can be used to yield fault tolerant estimates of such systems is the interacting multiple model (IMM) filtering algorithm, which is known to provide very accurate results. However, having been derived for a general class of systems with switching parameters, the IMM filter does not utilize the independence of the measurement errors in different channels, nor does it exploit the fact that the process noise is of low intensity. Thus, the implementation of the IMM in this case is computationally expensive. A new estimation technique is proposed herein, that explicitly utilizes the aforementioned properties. In the resulting estimation scheme separate measurement channels are handled separately, thus reducing the computational complexity. It is shown that, whereas the IMM complexity is exponential in the number of fault-prone measurements, the complexity of the proposed technique is polynomial. A simulation study involving spacecraft attitude estimation is carried out. This study shows that the proposed technique closely approximates the full-blown IMM algorithm, while requiring only a modest fraction of the computational cost.

In this paper, we propose an efficient, accurate, and stable algorithm for the numerical calculation of Carson's integral. The normalization of physical and electrical variables makes the proposed algorithm suitable for broad range evaluations of overhead line ground impedances. This allows the user to handle any practical line configuration with any electrical parameters. The algorithm uses, based on trapezoidal integration, a truncation criterion and identification of the first zero-crossing of the integrand as fundamental parts. Comparisons, analysis of error and computational efficiency between the proposed algorithm and an adaptive Gauss/Lobatto quadrature rule are provided. Index Terms--Carson ground return impedance, electromagnetic transient analysis, skin effect, transmission line.

Transient currents are quite common in power systems. They can be presented for both faults and switching events. In the case of switching events, such transients should not cause an overcurrent relay operation; therefore a dependable and secure relay response becomes a critical matter. Meanwhile, proper techniques must be used to prevent undesirable relay operations due to transient current and harmonics of current in the power system. This paper uses the concept of symmetrical components to discriminate fault from nonfault events. For this purpose, a criterion function is proposed using the above-mentioned components. The advantage of the proposed algorithm is shown by simulation of some typical transient currents due to transformer energizing and induction motor starting using EMTDC/PSCAD software. Index Terms--Fault, overcurrent relay, switching, symmetrical components.

Truncating finite difference time domain meshes using exact radiation boundary conditions is computationally expensive. The computational bottleneck stems from the global nature of the resulting boundary update scheme, which calls for the evaluation of retarded-time boundary integrals at each time step. The classical evaluation of such integrals requires O([N.sup.2.sub.s]) operations per time step where [N.sub.s] is the number of spatial field sampling points on the boundary. Here, the plane wave time domain algorithm is used to evaluate retarded-time boundary integrals in O ([N.sub.s] [log.sup.2] [N.sub.s]) operations per time step and thereby accelerate finite difference time domain solvers that impose exact radiation boundary conditions. The effectiveness of the resulting approach and its computational complexity are demonstrated through several numerical examples. Index Terms--Boundary integrals, exact global boundary conditions, finite-difference time-domain (FDTD), Plane wave time domain algorithm.

We present a fast solver for the Helmholtz equation on long, thin structures. It operates on an integral equation formulation of the problem, in which the solution is represented as a superposition of fields generated by sources on the structure (usually on the boundary or boundaries of the structure). It uses a standard iterative solver for linear equations, in conjunction with a novel method for applying the forward matrix, whose computational complexity is O(N), where N is the number of points on which the integral equation is solved. The algorithm is suitable for structures in either two dimensions (2-D) or three dimensions. It does not depend in any great detail on the specifics of the Helmholtz equation, and, thus, is also suitable for similar equations. We demonstrate the algorithm by using it to simulate scattering in 2-D from dielectric structures, using an integral equation formulation constructed using a combination of single-layer and double-layer potentials, yielding a second-kind integral equation. Numerical results show the algorithm to be efficient and accurate. Index Terms--Fast algorithm, Helmholtz equation, one-dimensional (1-D) fast multipole method (FMM), scattering.

The multipath equalization time delay estimator (METDE) provides an adaptive approach for estimating the difference in arrival times of a signal received at spatially separated sensors as well as the multipath channel characteristics. However, the parameter estimates of the METDE are biased in the presence of noise. The METDE algorithm is improved for unbiased parameter estimation via minimizing a modified cost function. Convergence behaviors and variances of the system variables of the amended method are derived. Computer simulations are included to corroborate the theoretical analysis and to evaluate the adaptive multipath delay estimation performance of the proposed algorithm.

In a large-scale information system such as a digital library or the web, a set of distributed caches can improve their effectiveness by coordinating their data placement decisions. Using simulation, we examine three practical cooperative placement algorithms, including one that is provably close to optimal, and we compare these algorithms to the optimal placement algorithm and several cooperative and noncooperative replacement algorithms. We draw five conclusions from these experiments: 1) Cooperative placement can significantly improve performance compared to local replacement algorithms, particularly when the size of individual caches is limited compared to the universe of objects; 2) although the amortizing placement algorithm is only guaranteed to be within 14 times the optimal, in practice it seems to provide an excellent approximation of the optimal; 3) in a cooperative caching scenario, the recent greedy-dual local replacement algorithm performs much better than the other local replacement algorithms; 4) our hierarchical-greedy-dual replacement algorithm yields further improvements over the greedy-dual algorithm especially when there are idle caches in the system; and 5) a key challenge to coordinated placement algorithms is generating good predictions of access patterns based on past accesses. Index Terms--Cache, cooperative, distributed, hierarchical, placement, replacement, web.

We consider the problem of finding a set of attribute values that give a high quality binary segmentation of a database. The quality of a segmentation is defined by an objective function suitable for the user's objective, such as 'mean squared error,' 'mutual information,' or '[chi square] each of which is defined in terms of the distribution of a given target attribute. Our goal is to find value groups on a given conditional domain that split databases into two segments, optimizing the value of an objective function. Though the problem is intractable for general objective functions, there are feasible algorithms for finding high quality binary segmentations when the objective function is convex, and we prove that the typical criteria mentioned above are all convex. We propose two practical algorithms, based on computational geometry techniques, which find a much better value group than conventional heuristics. Index Terms--Value groups, binary segmentation, categorical test, decision tree, data reduction, data mining.

Generating abductive explanations is the basis of several problem solving activities such as diagnosis, planning, and interpretation. Temporal abduction means generating explanations that do not only account for the presence of observations, but also for temporal information on them, based on temporal knowledge in the domain theory. We focus on the case where such a theory contains temporal constraints that are required to be consistent with temporal information on observations. The aim of this paper is to propose efficient algorithms for computing temporal abductive explanations. Temporal constraints in the theory and in the observations can be used actively by an abductive reasoner in order to prune inconsistent candidate explanations at an early stage during their generation. However, checking temporal constraint satisfaction frequently generates some overhead. In the paper, we analyze two incremental ways of making this process efficient. First we show how, using a specific class of temporal constraints (which is expressive enough for many applications), such an overhead can be reduced significantly, yet preserving a full pruning power. In general, the approach does not affect the asymptotic complexity of the problem, but it provides significant advantages in practical cases. We also show that, for some special classes of theories, the asymptotic complexity is also reduced. We then show how, compiled knowledge based on temporal information, can be used to further improve the computation, thus, extending to the temporal framework previous results in the case of atemporal abduction. The paper provides both analytic and experimental evaluations of the computational advantages provided by our approaches. Index Terms--Abductive reasoning, temporal reasoning, computing explanations, efficient algorithms, knowledge-based systems, knowledge compilation.

A gamma-ray tracking algorithm has been implemented and tested, using simulated data, for gamma rays with energies between 0.1 and 2 MeV, and its performance evaluated for a 90-mm-long, 60-mm-diameter, cylindrical, 36 (6 x 6) segment detector. The performance of the algorithm in two areas was determined: Compton suppression and Doppler shift correction. It was found that for gamma rays of energies around 1 MeV, a ratio of photopeak counts to total counts of 2:3 could be obtained using the tracking algorithm, with only a 2% reduction in detection efficiency, compared to the untracked data. Approximately 80% of first interaction points could be correctly identified, enabling a good Doppler shift correction. A detector of the type simulated has recently been delivered, together with a compactPCI digital data acquisition system comprising 36 12-bit, 40-MHz flash ADCs, and 6 200-MHz DSPs. Some initial data has been recorded using this system, and the performance of the tracking algorithm on this real data is comparable to its performance on simulated data.

Undesirable but unavoidable natural events or human errors will occur to disrupt normal power system operation. Engineers must apply protective relay systems for the most probable events based on practical judgment to minimize service interruptions and damage to the equipment. Understanding the performance and limitations of different protective relay systems is vital in their application. One can explain the philosophy of system protection, the basic theory of a protective relay system, and the relay algorithms from classroom sessions and/or computer simulation. However, it is difficult, if not impossible, to evaluate the actual performance and possible misoperation of a relay in the field just through computer simulation or benchtop testing. This paper introduces the usage of the Power System Simulation Laboratory at the Energy Systems Research Center to enhance the teaching and research activities in relay education. A brief description of the laboratory configuration and the design of the relay testing station are presented. Some sample examples to illustrate the use of this facility for undergraduate and graduate education in protective relay systems are also included in this paper. Index Terms--PC-based dynamic testing station, physical laboratory, protective relay testing and algorithm evaluation.

In this paper, we propose a fast multi-resolution block-matching algorithm (BMA) using multiple motion vector (MV) candidates and spatial correlation in MV fields, called a multi-resolution motion search algorithm (MRMCS). The proposed MRMCS satisfies high estimation performance and efficient LSI implementation. This paper describes the MRMCS with three resolution levels. At the coarsest level, two MV candidates are obtained on the basis of minimum matching error for the next search level. At the middle level, the two candidates selected at the coarsest level and the other one based on spatial MV correlation at the finest level are used as center points for local searches, and a MV candidate is chosen for the next search level. Then, at the finest level, the final MV is obtained from local search around the single candidate obtained at the middle level. This paper also describes an efficient LSI architecture based on the proposed algorithm for low bit-rate video coding. Since this architecture requires small number of processing elements (PEs) and a small size on-chip memory, MRMCS can be implemented with a much smaller number of gates than other conventional architectures for full-search BMA while keeping a negligible degradation of coding performance. Moreover, the proposed motion estimator can support an advanced prediction mode (8 x 8 prediction mode) for H.263 and MPEG-4 video encoding. We implement this architecture with about 25K gates and 288 bytes of RAM for a search range of [-16.0, + 15.5] by using a synthesizable VHDL. Index Terms--H.263, LSI, motion estimation, MPEG-4, multiresolution, spatial correlation.

By describing 3-D images as ellipsoids, the authors propose that they can utilize alpha-trimmed mean statistics to classify 3-D images. A discussion of the use of the Hough transform algorithm is presented.

A novel pseudo-noise resampling method is proposed to mitigate the impact of outliers in Unitary ESPRIT. The method does not require any additional data in that it exploits exactly the same data snapshots as the failed estimate itself. It uses synthetically generated data as in the bootstrap technique instead of unavailable additional data. The technique uses the main idea of the estimator bank approach. Another major difference is that the standard estimator bank approach requires a priori knowledge of source localization sectors, while the new technique is free of such a restriction.

Using phantom and human subject data, we have compared a number of reconstruction algorithms to the current 'gold standard' method, 3D Reprojection Method (3DRP), in an effort to validate them as practical 3D reconstruction alternatives for dynamic PET scanning. The algorithms evaluated were (a) Fourier Rebinning (FORE) followed by 2D Filtered Back Projection (2D FBP), (b) FORE followed by Ordered Subsets Expectation-Maximization (OSEM) and (c) FORE followed by Space Alternating Generalized Expectation-Maximization (SAGE). The main benefits of these methods are two-fold: significantly shorter reconstruction times and improved signal-to-noise ratio at matched resolution for the iterative schemes. We demonstrate that FORE + 2D FBP can replace 3DRP with no significant impact on subsequent data analysis. In particular, distribution volume ratios (DVRs) obtained with FORE + 2D FBP differed from those obtained with 3DRP by 0.3 [+ or -] 0.7% (n = 33) and -0.7 [+ or -] 1.1% (n = 27) for cortical and cerebellar input functions, respectively. Both OSEM and SAGE performed well compared to 3DRP, with improved noise characteristics and DVR differences of < 3%.

We report the study of a second-order [Sigma]-[Delta] modulator for audio applications with a nine-levels quantizer. The feedback digital-to-analog has been implemented with four capacitors and two voltage references, taking the structure typical of fully differential systems. We first discuss the performance achievable with the most popular dynamic element-matching techniques, namely the data-weighted averaging and individual-level averaging. We then introduce a new algorithm, the double-index averaging, that performs better than the other algorithms by 10-15 dB over the entire dynamic range of the modulator, and 6 dB at the peak of signal/(quantization noise + total harmonic distortion). Index Terms - Analog-digital conversion, sigma-delta modulation.

This paper deals with the problem of robust adaptive array beamforming for cyclostationary signals. By exploiting the signal cyclostationarity, the SCORE algorithms presented in [6] have been shown to be effective in performing adaptive beamforming without requiring the direction vector of the desired signal. However, these algorithms suffer from severe performance degradation even if there is a small mismatch in the cycle frequency of the desired signal. In this paper, we first evaluate the performance of the SCORE algorithms in the presence of cycle frequency error (CFE). An analytical formula is derived to show the behavior of the performance degradation due to CFE. based on the theoretical result, we then develop an efficient method in conjunction with the SCORE algorithms to achieve robust adaptive beamforming against the CFE. Several simulation examples for confirming the theoretical analysis and showing the effectiveness of the proposed method are also presented. Index Terms - Adaptive arrays, beamforming, cyclostationary signals.

Substantial technological strides in the are of very large scale integration (VLSI) have enabled the possibility of building workable, cost-efficient parallel machines with tens of thousands of processors. The first known adaptive time- and VLSI-optimal sorting algorithm on a new, more powerful architecture called mesh with multiple broadcasting that is commercially available in the DAP family of multiple computers is proposed.

A blind maximum likelihood method for demodulating multiple co-channel digital signals received synchronously at an antenna array has been suggested. Two computationally efficient block algorithms for determining the array response and signal estimates: iterative least-squares with projection and iterative least-squares with enumeration. The performance of these algorithms are analyzed employing both analysis and simulation. Simulation results indicate that the detection error probability provides a good approximation to the performance of the blind algorithms.

The performances of recursive and nonrecursive algorithms are compared for a fast changing peak position by simulations. The recursive algorithms provide better response and accuracy of time delay estimation even when the sampling period decreases. The average magnitude difference function is employed as an application example. This study illustrates the advantages of utilizing a recursive correlation function in the feedback loop of a tracking system employing a single digital signal processor.

We consider the problem of mining association rules on a shared-nothing multiprocessor. We present three algorithms that explore a spectrum of trade-offs between computation, communication, memory usage, synchronization, and the use of problem-specific information. The best algorithm exhibits near perfect scaleup behavior, yet requires only minimal overhead compared to the current best serial algorithm. Index Terms - Data mining, association rules, parallel algorithms.

The nonhomogeneous normalized version of the delayed least mean square algorithm with a decreasing step size has sure convergence in a mixing input condition and when a certain law of large numbers is satisfied. In nonhomogeneous case, rigorous sample convergence analysis for the least mean square type algorithms, when they are made to operate with a delay in the coefficient adaption, is absent.

A large number of algorithms exist for digital control which work well with high precision floating point computation. When implemented in finite wordlength processors without inherent floating point capability, these control algorithms often fail to demonstrate their effectiveness. The software described here has been developed to test the performance of digital controllers under such finite wordlength situations. It is also shown, how, with a suitable restructuring of the conventional discrete-time shift operator models, considerably improved robustness to such situations can be achieved with marginally added computations. Key words: Controller implementation, fixed-point arithmetic, finite wordlength control.

The asymptotic performance of the second-order ESPRIT, fourth-order ESPRIT and virtual ESPRIT (VESPA) algorithhm was assessed. Specifically, the asymptotic variance of the direction-of-arrival estimates derived by the least squares versions of the algorithms was analyzed under the assumption of a non-Gaussian signal. Second-order ESPRIT was found to be the best estimator based on the results and Monte Carlo simulations.

Performance analysis of four parallel sorting algorithms using the LogP model shows that the model is a good predictor of communication performance. The LogP model uses communication latency, overhead, bandwidth and the number of processors as parameters. The analysis involves the implementation of the algorithms in Split-C, followed by comparison of the LogP-predicted performances and the actual ones on a CM-5 for a range of problem sizes. Results from the study suggest that overhead is the most important communication parameter in sorting algorithms.

It is well known that dc offsets degrade the performance of analog adaptive filters. In this paper, the effects of dc offsets on four variations of the stochastic gradient algorithm are analyzed. Assuming a Gaussian probability distribution for the input signal and error signal, the output mean squared error (MSE) performance in the presence of dc offsets is evaluated for each of the algorithms. The theoretical work is compared with computer simulations and the results, together with convergence properties of each of the algorithms and their respective hardware requirements, are used in selecting the most appropriate algorithm. Although a Gaussian input distribution is assumed, it may reasonably be inferred that the critical results obtained should also hold for other input distributions.

A common framework to compare and characterize phase aberration correction algorithms based on their stability, capacity to reduce phase errors and noise sensitivity shows that low interelement correlation of the transducer arrays causes defective correction from one row to another. Variations in the size and position of the aperture area that generates the reference signal induce changes in the performance of the algorithms. Algorithms that use large reference frames are less affected by missing elements and noise than those using small reference areas.

Sliding window order recursive least-squares (SW-ORLS) algorithms comprise of two independent features, global array architecture and local cell implementation. Global array architecture deals with the linking of basic modules forming SW-ORLS estimates of different orders while local cell implementation discusses implementation of one-step order update by various forms of time-recursive scalar SWLS estimation. Time and order updates of SW-ORLS are discussed as a function of two hyperbolic Householder Transformation algorithms.

Exact and first-order error analysis helps analyze the numerical properties of the fixed-point (FP) conditions of the Schur and split Schur algorithms that calculate reflection coefficients. The Schur algorithm error-weight vectors match Levinson-like recursions while those of split Schur propagation mechanism satisfy split-Levinson-like recursions. Schur algorithm gives better results for signals of small eigenvalue spread than the split Schur under FP conditions.

Two distinct algorithms associated with packet filtering were studied to elucidate the bridges that interconnected Ethernet's local area network (LAN) to Fiber Distributed Data Interface (FDDI) backbones. The algorithm performance was compared with maximum Ethernet traffic and the rise in traffic generated on a local Ethernet. FDDI served as a backbone for lower-speed LAN's such as Token Ring and Ethernet.

A comparative study of the Rosenblatt, Backpropagation and LMS training algorithm, with respect to a single layer perceptron, revealed the three algorithms to exhibit negligible learning performance differences when selecting specific parametric values. The perceptron demonstrated an ability to recognize a specific nonlinear system possessing Gaussian inputs. A memoryless nonlinear device facilitated the output processing of the perceptron which was an adaptive linear combiner.

A comparison of the accuracy of several orthogonal-grid finite-difference time-domain (FDTD) schemes is made in both two and three dimensions. The relative accuracy is determined from the dispersion error associated with each algorithm and the number of floating point operations required to obtain a desired accuracy level. In general, in both 2-D and 3-D, fourth order algorithms are more efficient than second order schemes in terms of minimizing the number of computations for a given accuracy level. In addition, in 2-D, a new second order approach proposed by Chen, Ney, and Hoefer is much more accurate than the popular Yee scheme for a given amount of computation and can be as efficient as the fourth order algorithms. In 3-D, Yee's algorithm is slightly more efficient than Chen's approach in terms of operations, but much more efficient in terms of memory requirements.

The supervised training of feedforward neural networks is often based on the error backpropagation algorithm. Our main purpose is to consider the successive layers of a feedforward neural network as the stages of a pipeline which is used to improve the efficiency of the parallel algorithm. A simple placement rule will be presented in order to take advantage of simultaneous executions of the calculations on each layer of the network. The analytic expressions show that the parallelization is efficient. Moreover, they indicate that the performances of this implementation are almost independent of the neural network architecture. Their simplicity assures easy prediction of learning performance on a parallel machine for any neural network architecture. The experimental results are in agreement with analytical estimates.

In this paper, two alternate formulations of the Yee algorithm, namely, the finite-difference time-domain (FDTD) vector-wave algorithm and the FDTD scalar-wave algorithm are examined and compared to determine their relative merits and computational efficiency. By using the central-difference divergence relation the conventional Yee algorithm is rewritten as a hybrid Yee/FDTD scalar-wave algorithm. It is found that this can reduce the computation time for many 3-D open geometries, in particular planar structures, by approximately two times as well as reduce the computer-memory requirements by approximately one-third. Moreover, it is demonstrated both mathematically and verified by numerical simulation of a coplanar strip transmission line that this hybrid algorithm is entirely equivalent to the Yee algorithm. In addition, an alternate but mathematically equivalent reformulation of the Enquist-Majda absorbing boundary condition based on the normal field component (relative to the absorbing boundary wall) is given to increase the efficiency of the hybrid algorithm in the modeling of open region problems. Numerical results generated by the hybrid Yee/scalar-wave algorithm for the Vivaldi antenna are given and compared with published experimental work.