/pdf/neural-network-toolbox-user-s-guide-ll1bn7nko0.pdf

Neural Network Toolbox™ User's Guide

A modelforalarge network of"neurons" withagraded response (orsigmoid input-output relation) is studied. Thisdeterministic system hascollective properties in veryclose correspondence withtheearlier stochastic model based onMcCulloch-Pitts neurons. Thecontent-addressable memoryandother emergent collective properties oftheorigi- nalmodelalso arepresent inthegraded response model. The idea that suchcollective properties areusedinbiological sys- temsisgiven addedcredence bythecontinued presence ofsuch properties formorenearly biological "neurons." Collective analog electrical circuits ofthekinddescribed will certainly function. Thecollective states ofthetwomodels haveasimple correspondence. Theoriginal modelwill continue tobeuseful forsimulations, because its connection tograded response sys- temsisestablished. Equations that include theeffect ofaction potentials inthegraded response system arealsodeveloped. Recent papers (1-3) haveexplored theability ofasystem of highly interconnected "neurons" tohaveuseful collective computational properties. Theseproperties emergesponta- neously inasystem having alarge numberofelementary "neurons." Content-addressable memory(CAM)isoneof thesimplest collective properties ofsucha system. The mathematical modeling hasbeenbasedon"neurons" that aredifferent bothfromreal biological neurons andfromthe realistic functioning ofsimple electronic circuits. Someof these differences aremajor enough thatneurobiologists and circuit engineers alike havequestioned whether real neural orelectrical circuits wouldactually exhibit thekindofbe- haviors foundinthemodelsystem evenifthe"neurons" wereconnected inthefashion envisioned. Twomajor divergences between themodelandbiological orphysical systems stand out.Realneurons (andreal physi- caldevices suchasoperational amplifiers that might mimic them) havecontinuous input-output relations. (Action po- tentials areomitted until Discussion.) Theoriginal modeling usedtwo-state McCulloch-Pitts (4)threshold devices having outputs of0or1only. Realneurons andreal physical circuits haveintegrative timedelays duetocapacitance, andthetime evolution ofthestate ofsuchsystems should berepresented byadifferential equation (perhaps withaddednoise). The original modeling usedastochastic algorithm involving sud- den0-1or1-0changes ofstates ofneurons atrandom times. Thispaper showsthat theimportant properties oftheorigi- nalmodelremain intact whenthese twosimplifications of themodeling areeliminated. Although itisuncertain wheth- ertheproperties ofthese newcontinuous "neurons" areyet close enough totheessential properties ofrealneurons (and/or their dendritic arborization) tobedirectly applicable toneurobiology, amajor conceptual obstacle hasbeenelimi- nated. Itiscertain that aCAM constructed onthebasic ideas

Neurons withgraded response havecollective computational properties likethoseoftwo-state neurons

Switched dynamical systems can be found in many practical electronic circuits, such as various kinds of power converters, chaos generators, etc. This paper is concerned with the filter design problem for a class of switched system with average dwell time switching. Mode-dependent full-order filters are designed taking a more practical phenomenon, the asynchronous switching into account, where “asynchronous” means that the switching of the filters to be designed has a lag to the switching of the system modes. New results on the stability and l2-gain analyses for the systems are first given where the Lyapunov-like functions during the running time of subsystems are allowed to increase. In light of the proposed Lyapunov-like functions, the desired mode-dependent filters can be designed in that the unmatched filters are allowed to perform in the interval of the asynchronous switching before the matched ones are applied. In H∞ sense, the problem of asynchronous filtering for the underlying systems in linear cases is formulated and the conditions of the existence of admissible asynchronous filters are obtained. Two examples are provided to show the potential of the developed results.

Asynchronous Filtering of Discrete-Time Switched Linear Systems With Average Dwell Time

Since the early 1990s, a large number of chaos-based communication systems have been proposed exploiting the properties of chaotic waveforms. The motivation lies in the significant advantages provided by this class of non-linear signals. For this aim, many communication schemes and applications have been specially designed for chaos-based communication systems where energy, data rate, and synchronization awareness are considered in most designs. Recently, the major focus, however, has been given to the non-coherent chaos-based systems to benefit from the advantages of chaotic signals and non-coherent detection and to avoid the use of chaotic synchronization, which suffers from weak performance in the presence of additive noise. This paper presents a comprehensive survey of the entire wireless radio frequency chaos-based communication systems. First, it outlines the challenges of chaos implementations and synchronization methods, followed by comprehensive literature review and analysis of chaos-based coherent techniques and their applications. In the second part of the survey, we offer a taxonomy of the current literature by focusing on non-coherent detection methods. For each modulation class, this paper categorizes different transmission techniques by elaborating on its modulation, receiver type, data rate, complexity, energy efficiency, multiple access scheme, and performance. In addition, this survey reports on the analysis of tradeoff between different chaos-based communication systems. Finally, several concluding remarks are discussed.

Wireless Chaos-Based Communication Systems: A Comprehensive Survey

This paper presents the first ever experimental demonstration of a multiple-input multiple-output (MIMO) visible light communications system employing four silicon (Si) light emitting diodes (LEDs) and four organic photodetectors (OPDs) as transmitters and receivers, respectively. The proposed link is relatively low cost and it employs the on-off keying (OOK) modulation format offering a data rate of 200 kb/s without the need for equalization, which is a significant increase compared with previous non-equalized systems. In order to speed up date rates further, we implement an artificial neural network (ANN) to classify the signal and correct the error induced by the matrix inversion at the receiver, allowing a gross bit rate of 1.8 Mb/s in the best case.

A MIMO-ANN system for increasing data rates in organic visible light communications systems

In this letter, a new class of cubature Kalman filters (CKF) based on a spherical simplex-radial rule is proposed to further improve accuracy and efficiency of the traditional CKF. The transformation group of the regular simplex and the moment matching method are utilized to compute the spherical and radial integrals, respectively. In addition, with the increase of the state dimension, the new CKF of the fifth degree uses fewer quadrature points than the traditional CKF with the same degree. Simulations on moments calculation and nonlinear state estimation show that the proposed CKFs can achieve higher accuracy and better efficiency than the traditional CKFs.

Spherical Simplex-Radial Cubature Kalman Filter

In this paper, the problem of source extraction in bandwidth-constrained wireless sensor networks is considered. The sensor observations are assumed to be instantaneous linear mixtures of sources in a sensing field, possibly corrupted by noise. Because of the bandwidth constraint, the observations are quantized before being transmitted. Three kinds of sensor network topologies (cluster-based sensor network, sensor network with a fusion center, and concatenated sensor network) are considered in order to show the impact of topology on the performance and energy efficiency. The mixtures are reconstructed based on the received quantized data, and source extraction is performed by using a fast fixed-point algorithm with prewhitening. This algorithm was proposed by Hyvarinen and Oja, published in Neural Computation, and has the advantage of simplicity and fast convergence. The case of source extraction where the sensed data are undistorted is used as the performance benchmark. The results show that in all these sensor networks, the performance can be close to that of the benchmarking case. The energy efficiency and lifetime of these sensor networks are compared when the same source extraction task is executed. The effects of sensor failure and link failure on the performance are also discussed.

Impact of Topology on Performance and Energy Efficiency in Wireless Sensor Networks for Source Extraction

This work addresses the channel-distortion problem and proposes a technique for channel equalization in chaos-based communication systems. The proposed equalization is realized by a modified recurrent neural network incorporating a specific training (equalizing) algorithm.

/pdf/a-neural-network-based-channel-equalization-strategy-for-cuusjv8sko.pdf

A neural-network-based channel-equalization strategy for chaos-based communication systems

The kernel method has been successfully applied to nonlinear adaptive filtering This brief presents a kernel affine projection sign algorithm (KAPSA), which is a nonlinear extension of the affine projection sign algorithm (APSA) The proposed KAPSA combines the benefits of the kernel method and the APSA, and is robust against non-Gaussian impulse interference In order to further improve the filtering performance, a variable step-size adjustment is incorporated into the KAPSA, resulting in a new variable step-size kernel APSA (VSS-KAPSA) without increasing the computational burden Simulations in the context of time-series prediction show that both the KAPSA and the VSS-KAPSA are robust against impulse interference and that both outperform other affine projection algorithms in terms of steady-state mean square errors

Kernel Affine Projection Sign Algorithms for Combating Impulse Interference

This paper studies a multi-cluster sensor network which is applied for source extraction in a sensing field. Both the performance of source extraction and the total energy consumption in the sensor network are functions of the number of clusters. In this paper, we aim at finding the optimal number of clusters by minimizing the effective energy consumption which is defined as the ratio of the performance of source extraction to the total energy consumption in the sensor network. A particle swarm optimization (PSO) algorithm is employed to form the clusters which enables every cluster to perform source extraction. The existence and the uniqueness of the optimum number of clusters is proven theoretically and shown by numerical simulations. The relationship between the optimum number of clusters and the various system parameters are investigated. A tradeoff between the performance and the total energy consumption is illustrated. The results show that the performance is greatly improved by adopting the multi-cluster structure of the sensor network.

Jiuchao Feng

Papers

Spherical Simplex-Radial Cubature Kalman Filter

Impact of Topology on Performance and Energy Efficiency in Wireless Sensor Networks for Source Extraction

A neural-network-based channel-equalization strategy for chaos-based communication systems

Kernel Affine Projection Sign Algorithms for Combating Impulse Interference

Minimizing effective energy consumption in multi-cluster sensor networks for source extraction