SUMMARY The emergence of streaming multicore processors with multi‐SIMD (single‐instruction multiple‐data) architectures and ultra‐low power operation combined with real‐time compute and I/O reconfigurability opens unprecedented opportunities for executing sophisticated signal processing algorithms faster and within a much lower energy budget. Here, we present an unconventional Fast Fourier Transform (FFT) implementation scheme for the IBM Cell, named transverse vectorization. It is shown to outperform (both in terms of timing and GFLOP throughput) the fastest FFT results reported to date for the Cell in the open literature. We also provide the first results for multi‐FFT implementation and application on the novel, ultra‐low power Coherent Logix HyperX processor. Copyright © 2011 John Wiley & Sons, Ltd.

The emergence of streaming multicore processors with multi-SIMD architectures and ultra-low power operation combined with real-time compute and I/O reconfigurability opens unprecedented opportunities for executing sophisticated signal processing algorithms faster and within a much lower energy budget. Here, we present an unconventional FFT implementation scheme for the IBM Cell, named transverse vectorization. It is shown to outperform (both in terms of timing or GFLOP throughput) the fastest FFT results reported to date in the open literature.

We consider systems whose parameters satisfy certain easily computable physical laws. Each parameter is directly measured by a number of sensors, or estimated using measurements, or both. The measurement process may introduce both systematic and random errors which may then propagate into the estimates. Furthermore, the actual parameter values are not known since every parameter is measured or estimated, which makes the existing sample-based fusion methods inapplicable. We propose a fusion method for combining the measurements and estimators based on the least violation of physical laws that relate the parameters. Under fairly general smoothness and nonsmoothness conditions on the physical laws, we show the asymptotic convergence of our method and also derive distribution-free performance bounds based on finite samples. For suitable choices of the fuser classes, we show that for each parameter the fused estimate is probabilistically at least as good as its best measurement as well as best estimate. We illustrate the effectiveness of this method for a practical problem of fusing well-log data in methane hydrate exploration.

One of the most important features of artificial neural networks in emerging, brain-inspired, nanoarchitectural design is their inherent ability to perform massively parallel, nonlinear signal processing. When operating in a system-wide asynchronous regime, such networks may exhibit a phenomenon referred to as "computational chaos", which impedes the efficient retrieval of information usually stored in the system's attractors. We illustrate the emergence of computational chaos from fixed point and limit cycle attractors for node communication delays in a widely used neural network model. In particular, the complete Lyapunov spectrum associated with the network dynamics is computed, and conditions that prevent the emergence of chaos are derived.

The problem of computing a route for a mobile agent that incrementally fuses the data as it visits the nodes in a distributed sensor network is considered. The order of nodes visited along the route has a significant impact on the quality and cost of fused data, which, in turn, impacts the main objective of the sensor network, such as target classification or tracking. We present a simplified analytical model for a distributed sensor network and formulate the route computation problem in terms of maximizing an objective function, which is directly proportional to the received signal strength and inversely proportional to the path loss and energy consumption. We show this problem to be NP-complete and propose a genetic algorithm to compute an approximate solution by suitably employing a two-level encoding scheme and genetic operators tailored to the objective function. We present simulation results for networks with different node sizes and sensor distributions, which demonstrate the superior performance of our algorithm over two existing heuristics, namely, local closest first and global closest first methods.

The image registration problem of finding a mapping that matches data from multiple cameras is computationally intensive. Current solutions to this problem tolerate Gaussian noise, but are unable to perform the underlying global optimization computation in real time. This paper expands these approaches to other noise models and proposes the Terminal Repeller Unconstrained Subenergy Tunneling (TRUST) method, originally introduced by B.C. Cetin et al. (1993), as an appropriate global optimization method for image registration. TRUST avoids local minima entrapment, without resorting to exhaustive search by using subenergy-tunneling and terminal repellers. The TRUST method applied to the registration problem shows good convergence results to the global minimum. Experimental results show TRUST to be more computationally efficient than either tabu search or genetic algorithms.

Abstract A fundamental aspect of content‐based image retrieval (CBIR) is the extraction and the representation of a visual feature that is an effective discriminant between pairs of images. Among the many visual features that have been studied, the distribution of color pixels in an image is the most common visual feature studied. The standard representation of color for content‐based indexing in image databases is the color histogram. Vector‐based distance functions are used to compute the similarity between two images as the distance between points in the color histogram space. This paper proposes an alternative real valued representation of color based on the information theoretic concept of entropy. A theoretical presentation of image entropy is accompanied by a practical description of the merits and limitations of image entropy compared to color histograms. Specifically, the L1norm for color histograms is shown to provide an upper bound on the difference between image entropy values. Our initial results suggest that image entropy is a promising approach to image description and representation.

A feedforward multilayer neural net is trained to learn the correspondence between seismic data and well logs. The introduction of a virtual input layer, connected to the nominal input layer through a special nonlinear transfer function, enables ultrafast (single iteration), near-optimal training of the net using numerical algebraic techniques. A unique computer code, named DeepNet, has been developed, that has achieved, in actual field demonstrations, results unattainable to date with industry standard tools.

Abstract The paper presents the implementation of a new class of massively parallel algorithms for solving certain time‐dependent partial differential equations (PDEs) on massively parallel supercomputers. Such PDEs are usually solved numerically, by discretization in time and space, and by applying a time‐stepping procedure to data and algorithms potentially parallelized in the spatial domain. In a radical departure from such a strictly sequential temporal paradigm, we have developed a concept of time‐parallel algorithms, which allows the marching in time to be fully parallelized. This is achieved by using a set of transformations based on eigenvalue‐eigenvector decomposition of the matrices involved in the discrete formalism. Our time‐parallel algorithms possess a highly decoupled structure, and can therefore be efficiently implemented on emerging, massively parallel, high‐performance supercomputers, with a minimum of communication and synchronization overhead. We have successfully carried out a proof‐of‐concept demonstration of the basic ideas using a two‐dimensional heat equation example implemented on the Intel Touchstone Delta supercomputer. Our results indicate that linear, and even superlinear, speed‐up can be achieved and maintained for a very large number of processor nodes.

Many space robotic systems would be required to operate in uncertain or even unknown environments. The problem of identifying such environment for compliance control is considered. In particular, neural networks are used for identifying environments that a robot establishes contact with. Both function approximation and parameter identification (with fixed nonlinear structure and unknown parameters) results are presented. The environment model structure considered is relevant to two space applications: cooperative execution of tasks by robots and astronauts, and sample acquisition during planetary exploration. Compliant motion experiments have been performed with a robotic arm, placed in contact with a single-degree-of-freedom electromechanical environment. In the experiments, desired contact forces are computed using a neural network, given a desired motion trajectory. Results of the control experiments performed on robot hardware are described and discussed.>

A method for the global configuration resolution for kinematically redundant manipulators is presented. This method is based on an efficient global optimization algorithm which uses a sub-energy tunneling function and terminal repellers. This optimization algorithm is reviewed, and its specialization to redundancy resolution is developed. Applications of this method and comparisons to null-space projection are presented.>

A type of dynamical system for solving optimization problems is introduced. The approach exploits a novel paradigm in nonlinear dynamics that is based upon the concept of terminal attractors and repellers. A class of dynamical systems-the unpredictable systems-is introduced and analyzed. These systems are represented in the form of coupled activation and learning dynamical equations whose ability to be spontaneously activated is based upon two pathological characteristics: (1) Such systems have zero Jacobian. As a result, they have an infinite number of equilibrium points which occupy curves, surfaces, or hypersurfaces. (2) At all of these equilibrium points, the Lipschitz condition fails, so the equilibrium points become terminal attractors or repellers, depending upon the sign of the periodic excitation. These characteristics result in multichoice response and lead to unpredictable dynamical systems. The systems can be controlled by sign strings which uniquely define the systems' behavior by specifying the direction of the motions at the critical points. By changing the combinations of signs in the code strings, a system can reproduce any prescribed behavior to a prescribed accuracy, which is why the unpredictable systems driven by sign strings are extremely flexible and can be exploited for solving optimization problems

The successful development of fifth-generation systems requires enormous computational capability and flexibility, necessitating the ability to achieve operational responses in hard real-time through optimal resource utilisation and introduction of adaptive control. This necessitates dynamically balancing the computational load among all the processing nodes in the system. In this paper we propose a graph-theoretic, receiver-initiated, distributed protocol for dynamic load balancing protocol in large-scale hypercube ensembles. Using attributed hypergraphs as the primary data structure for constraint modelling and dynamic optimisation, we consider systems running precedence-constrained heterogeneous tasks. Fault Tolerance is ensured by incorporating a dynamic integrity check for the decision nodes and their subsequent re-election if needed. Simulation studies are used to analyse the algorithm performance and correctness.

The stability, capacity, and design of a nonlinear continuous neural network are analyzed. Sufficient conditions for existence and asymptotic stability of the network's equilibria are reduced to a set of piecewise-linear inequality relations that can be solved by a feedforward binary network, or by methods such as Fourier elimination. The stability and capacity of the network is characterized by the postsynaptic firing rate function. An N-neuron network with sigmoidal firing function is shown to have up to 3/sup N/ equilibrium points. This offers a higher capacity than the (0.1-0.2)N obtained in the binary Hopfield network. It is shown that by a proper selection of the postsynaptic firing rate function, one can significantly extend the capacity storage of the network.>