Large Number Of Inputs Research Articles

Grid-based switch fabrics: a new approach in designing fault-tolerant ATM switches

ATM is the switching and multiplexing technology chosen to be used in the implementation of B-ISDN, because of its superiority in fast packet switching. However, the use of ATM switches with large number of input and output ports have been proven to be a bottleneck in wide area ATM networks. In this paper, we propose a new space-division grid-based ATM architecture with fault tolerant characteristics and minimal number of switching elements (SE's).

Interspike interval variability for balanced networks with reversal potentials for large numbers of inputs

The hypothesis that the variability in the discharge of cortical neurons results from balanced excitation and inhibition is analyzed. A method is presented for analyzing the integrate and fire neural model with reversal potentials which enables the interspike interval distribution to be calculated in the Gaussian approximation. Results are presented for the perfect integrator model, for which a stable value of the membrane potential in the absence of the spiking mechanism exists. The results show close agreement with numerical simulations for large numbers of small amplitude inputs. The coefficient of variation is consistently less than 1.0, as observed in cortical neurons.

Robustness margin computation for large scale systems

Large scale systems have large numbers of inputs and outputs, and include whole chemical plants as well as some unit operations, such as paper machines, polymer film extruders, and adhesive coaters. The importance of ensuring robustness of the closed loop system to model uncertainties increases as the process dimensionality increases; hence developing algorithms for computing robustness margins for large scale systems is of immense practical importance. Computational complexity is a tool of computer scientists which has had impact in understanding large scale optimization problems, both theoretically and in terms of finding computational solutions. Computational complexity theory is used to determine the level of accuracy and computational speed that are obtainable by algorithms for computing robustness margins, and as to which algorithms are likely for providing practical robustness margin computation for large scale systems.

The efficiency of neuron-MOS transistors in threshold logic

Using MOS-transistors with floating gate (Neuron MOS or νMOS) for building threshold logic is discussed. Two ways of νMOS threshold logic implimention – static and clocked – are under consideration. Methodology of νMOS circuit design is given. Majority voting gate (MVG) is used as an example of threshold gate with worst conditions for getting a large number of inputs. The possibility of implementing a MVG with a certain number of inputs is the possibility of building a threshold gate with a threshold alterable in real time (from OR to AND-function) with the sum of input weights equal to the number of MVG inputs. The maximum number of threshold gate inputs is estimated depending upon the deviations of the elements dimensions and parameters inside the chip. It is shown that it is difficult to implement a static νMOS MVG with a number of inputs more than 10. For the same conditions, the number of inputs of clocked νMOS MVG is as large as many tens. A clocked νMOS threshold gate with alterable in real-time input weights and threshold is proposed. Delay time and chip area for such a circuits are estimated.

Modulation of cell firing in the nucleus accumbens.

Pennartz et al. have proposed that functions of the nucleus accumbens (NA) are subserved by the activity of ensembles of neurons rather than by an overall neuronal activation. Indeed, the NA is a site of convergence for a large number of inputs from limbic structures that may modulate the flow of prefrontal cortical information and contribute to defining such ensembles, as exemplified in the ability of hippocampal input to gate cortical throughput in the nucleus accumbens. NA neurons exhibit a bistable membrane potential, characterized by a very negative resting membrane potential (down state), periodically interrupted by plateau depolarizations (up state), during which the cells may fire in response to cortical inputs. A dynamic ensemble can be the result of a distributed set of neurons in their up state, determined by the moment-to-moment changes in the spatial distribution of hippocampal inputs responsible for transitions to the up state. Ensembles may change as an adaptation to the contextual information provided by the hippocampal input. Furthermore, for dynamic ensembles to be functionally relevant, the model calls for near synchronous transitions to the up state in a group of neurons. This can be accomplished by the cell-to-cell transfer of information via gap junctions, a mechanism that can allow for a transfer of slow electrical signals, including "up" events between coupled cells. Furthermore, gap junction permeability is tightly modulated by a number of factors, including levels of dopamine and nitric oxide, and cortical inputs, allowing for fine-tuning of this synchronization of up events. The continuous selection of such dynamic ensembles in the NA may be disputed in schizophrenia, resulting in an inappropriate level of activity of thalamocortical systems.

Assessing the compensation for electricity wayleaves

This paper examines the measure of compensation payable for electricity wayleaves through afforestation and considers the merits of using a discounted cash flow approach to calculate the level of compensation. The research finds that the majority of negotiations take place “under the shadow” of compulsory powers and that the measure of compensation is based on the principle of equivalence; the compensation payable being the loss to the landowner and not the gain to the electricity supply company. It concludes that great caution is needed when applying a discounted cash flow approach to the valuation of afforestation owing to the large number of inputs and the sensitivity of the estimated land value to changes to key variables such as the discount rate and tree species.

Reducing on-line computational demands in model predictive control by approximating QP constraints

Abstract In this paper, we propose two Model Predictive Control algorithms, whose on-line computational demands are significantly smaller than that for conventional Model Predictive Control algorithms, for control of large-scale constrained linear systems. We show that closed-loop stability can be guaranteed under some conditions. We also propose an optimal anti-windup scheme for approximating Model Predictive Control (thus eliminating the need for solving an on-line optimization problem) and derive a quantitative condition under which Model Predictive Control can be approximated effectively. These results make Model Predictive Control a very attractive candidate to be applied to systems with small sampling times and/or with a large number of inputs, and address achievable constrained performance by any anti-windup design.

On robust nonlinear modeling of a complex process with large number of inputs using m-QRcp factorization and C/sub p/ statistic

The problem of modeling complex processes with a large number of inputs is addressed. A new method is proposed for the optimization of the models in minimum C(p) statistic sense using QR with a modified scheme of column pivoting (m-QRcp) factorization. Two different classes of multilayer nonlinear modeling problems are explored: 1) in the first class of models, each layer comprises multiple linearly parameterized submodels or cells; the individual cells are optimally modeled using QR factorization, and m-QRcp factorization ensures optimal selection of variables across the layers. 2) The nonhomogeneous feed-forward neural network is chosen as the second class of models, where the network architecture and structure are optimized in terms of best set of hidden links (and nodes) using m-QPcp factorization. In both the cases, the optimization is shown to be direct and conclusive. The proposed is a generic approach to the optimal modeling of complex multilayered architectures, which leads to computationally fast and numerically robust parsimonious designs, free from collinearity problems. The method is largely free from heuristics and is amenable to automated modeling.

UNCERTAINTY ANALYSIS OF FOOD-CHAIN MODELS

Each sub-model has to be specified by a set of parameters, which are imperfectly known; it may also be influenced by unpredictable exogenous input. These uncertain inputs are collectively represented by the upward arrows in the diagram. The uncertainty about these inputs is often modelled by considering them as random variables. The random effects propagate through the chain, by which some accumulation will occur. Uncertainty analysis provides methods to study the consequences of these random effects. The goal of such analyses is to evaluate if the joint effect on the final output y is still acceptably small, and to pinpoint the sources that contribute most to the uncertainty about the final output y. The results may be used to assess the potential benefits of control measures and of more accurate estimation of model parameters. At present, most uncertainty analyses are performed computer-experimentally, by Monte Carlo methods. The model output is calculated for a representative sample of the uncertain model inputs. If the sample is large, the computational burden may be considerable. Several statistical methods have been developed to produce efficient Monte Carlo samples in order to keep the computational efforts as small as possible and to efficiently analyse the results. For a survey, see e.g. Helton (1993). In Section 1, uncertainty and uncertainty contributions are characterised in terms of variances and variance components. Section 2 treats standard analyses based on regression-type approximations of the model output, the uncertain inputs serving as regressors. The analysis, which yields uncertainty contributions per individual input, is valid if the regression-approximation is of good quality. The approximation may fail because of non-linearities, especially in combination with a large number of inputs. Moreover, one may be interested in contributions of groups of inputs rather than individual inputs. Section 3 describes how, in such cases, one can have recourse to methods that are not based on regression-approximations, but that are more computer-intensive. Sections 2 and 3 contain examples of uncertainty analyses of models at the beginning of the food-chain. The paper ends with a discussion section

Neural operation using stimulated Brillouin scattering in optical fiber

We investigate an optoelectronic system that can perform weighting and threshold operations similarly to a neuron. The weighted summation is carried out optically in an optical fiber by using stimulated Brillouin scattering. It is shown that magnitude and sign of a weight associated with an input can be set by changing the signal intensity and wavelength, respectively. A signal traveling in the backward direction is required to sense additions and subtractions. Due to inherent parallel nature of optical processing, this system can outperform a digital system for large numbers of inputs.

Nonlinear system input structure identification: two stage fuzzy curves and surfaces

When modeling a complex, poorly defined, nonlinear problem with hundreds of possible inputs, we must identify the significant inputs before any known nonlinear modeling techniques can be applied. We introduce two stage fuzzy curves and surfaces and use them to 1) automatically and quickly order large numbers of inputs according to their significance, 2) eliminate spurious inputs, and 3) eliminate inputs dependent on the significant inputs.

MULTIDIMENSIONAL REALIZATION OF LARGE SCALE UNCERTAIN SYSTEMS FOR MULTIVARIABLE STABILITY MARGIN COMPUTATION

SUMMARY The prevailing framework for robust stability and performance analysis requires that the uncertain system be written as a linear fractional transformation of the uncertain parameters. This problem is algebraically equivalent to the problem of deriving the state space realization for a multidimensional transfer function matrix, for which a systematic algorithm was recently provided by Cheng and DeMoor. 1 In this work an algorithm is developed that reduces the dimension of the realizations while improving numerical accuracy, reducing computational expense, and reducing run-time memory requirements. Such improvements are required for the realization of large scale uncertain systems, which have large numbers of inputs, outputs, states, and=or uncertain parameters.

MULTIDIMENSIONAL REALIZATION OF LARGE SCALE UNCERTAIN SYSTEMS FOR MULTIVARIABLE STABILITY MARGIN COMPUTATION

The prevailing framework for robust stability and performance analysis requires that the uncertain system be written as a linear fractional transformation of the uncertain parameters. This problem is algebraically equivalent to the problem of deriving the state space realization for a multidimensional transfer function matrix, for which a systematic algorithm was recently provided by Cheng and DeMoor. In this work an algorithm is developed that reduces the dimension of the realizations while improving numerical accuracy, reducing computational expense, and reducing run-time memory requirements. Such improvements are required for the realization of large scale uncertain systems, which have large numbers of inputs, outputs, states, and/or uncertain parameters. © 1997 by John Wiley & Sons, Ltd.

Multilayered fuzzy behavior fusion for real-time reactive control of systems with multiple sensors

Fuzzy linguistic rules provide an intuitive and powerful means for defining control behavior. Most applications that use fuzzy control feature a single layer of fuzzy inference, mapping a function from one or two inputs to equally few outputs. Highly complex systems, with large numbers of inputs, may also benefit from the use of qualitative linguistic rules if the control task is properly partitioned. This paper presents a modular fuzzy control architecture and inference engine that can be used to control complex systems. The control function is broken down into multiple local agents, each of which samples a subset of a large sensor input space. Additional fuzzy agents are employed to fuse the recommendations of the local agents. Real-time implementation without special hardware is possible by using singleton output values during fuzzy rule evaluation. A development tool is used to translate a fuzzy programming language offline for fast execution at run time. Using this system, a multilayered fuzzy behavior fusion based reactive control system has been implemented on an autonomous mobile robot, MARGE, with great success. MARGE won first place in Event III of the 1993 Robot Competition sponsored by the American Association for Artificial Intelligence.

Numerical Algorithms for Mapping Boundaries of Manipulator Workspaces

Numerical algorithms for mapping boundaries of manipulator workspaces are developed and illustrated. Analytical criteria for boundaries of workspaces for both manipulators having the same number of input and output coordinates and redundantly controlled manipulators with a larger number of inputs than outputs are well known, but reliable numerical methods for mapping them have not been presented. In this paper, a numerical method is first developed for finding an initial point on the boundary. From this point, a continuation method that accounts for simple and multiple bifurcation of one-dimensional solution curves is developed. Second order Taylor expansions are derived for finding tangents to solution curves at simple bifurcation points of continuation equations and for characterizing barriers to control of manipulators. A recently developed method for tangent calculation at multiple bifurcation points is employed. A planar redundantly controlled serial manipulator is analyzed, determining both the exterior boundary of the accessible output set and interior curves that represent local impediments to motion control. Using these methods, more complex planar and spatial Stewart platform manipulators are analyzed in a companion paper.

An alternative approach for generation of membership functions and fuzzy rules based on radial and cubic basis function networks

The theoretically attractive fact that the radial basis function networks can be interpreted as fuzzy systems is of small importance for practical applications such as diagnosis and quality control with large numbers of inputs or hidden neurons, due to the lack of transparency of the resulting fuzzy systems. A novel method for the generation of fuzzy classification systems based on radial basis function networks with restricted Coulomb energy learning is presented. The neural network and the learning algorithm are modified for easy hardware implementation by introducing cubic basis functions. The proposed methods are tested with three application examples. The simulation results show the generation of compact, transparent fuzzy classification systems with good performance.

An efficient technique for obtaining unate implementation of functions through input encoding

Unate circuits have the desirable property that a single stuck-at fault manifests as multiple unidirectional error at the output. If the valid outputs belong to a unidirectional error detecting code, we can add circuitry to determine if the outputs have suffered any error and implement a self-checking system. The key contribution here is that we can start with an arbitrary circuit with arbitrary number of inputs and outputs and convert that to a unate circuit with coded input outputs. Earlier techniques were based on enumeration of truth space and therefore suffered from a combinatorial explosion for large number of inputs. The technique presented here proceeds with local transformations, producing a sub-optimal solution as a consequence, as opposed to no solution. The results are shown to be within acceptable bounds.

Neural networks in designing fuzzy systems for real world applications

Abstract A special multilayer perceptron architecture known as FuNe I is successfully used for generating fuzzy systems for a number of real world applications. The FuNe I trained with supervised learning can be used to extract fuzzy rules from a given representative input/output data set. Furthermore, optimization of the knowledge base in possible including the tuning of membership functions. The new method employed to identify the rule relevant nodes before the rules are extracted makes FuNe I suitable for applications with large number of inputs. Some of the real world applications in areas of state identification and image classification show encouraging results in a shorter development time. Expert knowledge is not compulsory but can be included in the automatically extracted knowledge base. The generated fuzzy system can be implemented in hardware very easily. A flexible prototype board is developed with a FPGA chip in order to run applications with up to 128 inputs and 4 outputs in realtime (1.25 million rules per second).

Advanced multilayer metallization schemes with copper as interconnection metal

Advanced metallization schemes are needed to take advantage of the miniaturization of microelectronic devices which are performing at increasingly high speeds. The demands on metallization center around (a) the increased resistance with lower cross-sectional areas and longer interconnect lengths and (b) stability with the surroundings during processing and use under high current densities and thin film stresses. A threefold attack is being pursued to solve these problems, which also duplicate the issues in packaging of these fast chips with large numbers of inputs and outputs: first is to make use of copper as the interconnection metal; second is to use a multilevel metallization scheme; finally there is a need for a low dielectric constant dielectric. In this paper we present a review of progress made in addressing the first two schemes together with a brief discussion of the third. Copper, a heretofore undesired metal in silicon integrated circuits, seems to show promise, with appropriate processing constraints, of fulfilling the projected needs of ultra-large-scale and giga-scale integration and perhaps even of packaging.

Learning curves of the clipped Hebb rule for networks with binary weights

Networks with binary weights are very important from both the theoretical and practical points of view. In this paper, we investigate the clipped Hebb rule for learning different networks of non-overlapping binary perceptrons under uniform distribution. We calculate exactly its learning curves in the limit of a large number of inputs, where the average behaviour becomes the typical behaviour. This calculation is performed using only very simple counting arguments and the central limit theorem. The results indicate that the clipped Hebb rule does indeed learn this class of networks. In particular, the generalization rates converge extremely rapidly, often exponentially, to perfect generalization. These results are very encouraging given the simplicity of the learning rule. The analytic expression of the learning curves are in excellent agreement with the numerical simulations.