ABSTRACT In this paper, learning algorithm for a single multiplicative spiking neuron (MSN) is proposed and tested for various applications where a multilayer perceptron (MLP) neural network is conventionally used. It is found that a single MSN is sufficient for the applications that require a number of neurons in different hidden layers of a conventional neural network. Several benchmark and real-life problems of classification and function-approximation are illustrated. It is observed that by incorporating nonlinear synaptic interaction, threshold variability, and spiking phenomena, learning in artificial neural networks can be made more efficient. Keywords: multiplicative spiking neuron, learning, classification, function approximation. 1. INTRODUCTION Many researchers have proposed several neuron models for artificial neural networks. Although all these models were primarily inspired from biological neuron, there is still a gap between philosophies used in neuron models for neuroscience studies and those used for artificial neural networks (ANN). Some of these models exhibit a close correspondence with their biological counterparts while others do not. Freeman (Freeman, 1988) has pointed out that while brains and neural networks share certain structural features, such as, massive parallelism, biological networks solve complex problems easily and creatively, but existing neural networks do not. He discussed the issues related to the similarities and dissimilarities between biological and artificial neural systems of present days. The main focus in the development of a neuron model for artificial neural networks is not only its ability to represent biological activities with its maximum intricacy, but also some mathematical properties e.g.; its capability as a universal function approximator. However, it can be advantageous for artificial neural networks if we can bridge the gap between biology and mathematics by investigating the learning capabilities of biological neuron models for the applications of classification, time-series prediction, function approximation etc. Here, we used the
ABSTRACT Software agents offer a promise to change electronic commerce trading by helping traders to purchase products based on their interests and preferences. E-commerce systems are increasingly recognizing the importance of giving additional value to customers by providing customized transactional experiences. We believe that increased support for collaboration is a logical next step in the evolution of the e-commerce systems. The main goal here is to create a collaborative multi-agent based e-commerce framework that allows autonomy, pro-activity, and personalization including an intelligent mall agent, a seller agent, and a buyer agent with their independent profiles. The basic idea is to get a fully qualified broker that can intelligently identify the buyer’s needs based on standard parameters that are given to help alleviate the problems of finding interest items. The proposed framework aims at giving meaningful responses to meaningful requests and to delivering appropriate items to people who need it, when they need it, in a manner that meets their interests. To demonstrate the proposed framework a prototype is implemented and tested.
People have trusted in face-to-face interaction more than any other modes of interactions to develop relationships and as a result of which one of the most concerned problem that the electronic world is facing, is the lack of trust. For online trust development, study of customer behavior is must as customers, in present context, have become more fickle and product cycles are shortening. This research addresses the issues of trust development in Internet shopping and proposes a systematic methodology to develop trust through orientation of electronic customers. All these objectives have been accomplished in three steps; firstly, the factors that influence trust development in Internet shopping are identified and their association with electronic customers’ Internet shopping experience is established. Secondly, the electronic customer base is divided into five segments, using a-priori predictive method. The data collected about the electronic customers was mapped onto these segments, which resonate with the identified segments. Lastly, the factors, influencing the trust in Internet shopping, identified were combined with the segments of electronic customers to develop a segment specific electronic customer orientation methodology. In this study, it is shown that, not all the electronic customers want the same treatment from the online vendors, as their needs and desires are different and they expect more personal treatment.
This paper presents a low power driven synthesis framework for the unique class of nonregenerative Boolean Read-Once Functions (BROF). A two-pronged approach is adopted, where the satisfiability of the functionality is first ensured at the logic level on the basis of the proposed ‘hybrid synthesis method’. The resulting circuitry is contrasted with the reduced disjunctive normal form (DNF), resulting from standard two-level synthesis tool, ESPRESSO and conjunctive normal form (CNF) expression obtained via, the conventional Tabulation method. The gate level schematics are then translated into MOS transistor descriptions via, static CMOS and stacked CMOS implementation styles. Leakage Control Transistors (LCTs) are also inserted between the pull-up and pull-down network nodes, so as to minimize the overall power consumption of the digital logic circuits designed. Furthermore, the effect of transistor re-ordering on the delay of the resulting CMOS digital designs is also investigated. The adopted synthesis procedures are all evaluated based on a common Energy Delay Product (EDP) metric. The SPICE simulation results obtained for a 350nm TSMC CMOS process are promising, as it reports 41.5% savings in EDP, 10.5% reduction in power and 17.7% decrease in delay for the proposed method, on an average, over the best of conventional methods.
ABSTRACT The use of data mining by organizations has grown sizeably because thanks to Moore’s Law, over the last decade especially, organizations have been able to gather vast amounts of data during their operations. However, the results accruing from this usage of data mining on operational data has been mixed. Data mining is intended to be a non-trivial process of identifying interesting patterns, where interesting infers valid, understandable, novel, and potentially useful patterns; and yet too frequently, the results have yielded uninteresting (irrelevant, and/or obvious) patterns. To increase the confidence of decision makers in the interestingness of discovered patterns, some researchers believe in the incorporation of domain prior-knowledge into the data mining process. In this paper, we present a new design artifact, that uses the analytic hierarchy process (AHP) to conceptualize and structure domain prior-knowledge, thus capturing a broader essence of domain knowledge on which data mining can be applied. Our method is built and evaluated using best practice design science principles and guidelines. The evaluation of the artifact occurs within the domain of brain trauma intensive care. This particular paper focuses on the design and design components of our artifact.
In this paper, minimum jerk movement on the constrained sphere was studied by using both theoretical analysis and experimental investigation. Based on the constraint optimal principle, it was obtained that the trajectory of the minimum jerk movement on the surface of sphere should be a geodesic. Experiments used a robot especially designed for studies of human-machine interactions to evaluate movements on a virtual surface. Experiments revealed that the nervous system tries to learn and represent a constrained surface so it can move most efficiently (i.e. move along a geodesic) with practice, and that subjects can generalize this knowledge to nearby locations not yet visited.
Electronic government is a complex phenomenon which involves technical, organizational, institutional and environmental aspects. Researchers from different disciplines are increasingly finding that using multiple methods can help to deal with complexity and obtain more comprehensive explanations. This paper argues that multi-method approaches can be useful for egovernment research. A set of advantages and challenges to multi-method approaches are introduced and then used to frame a case analysis. Two case studies involving multi-method approaches to e-government research are presented to illustrate strategies for responding to implementation challenges in both large-scale and small-scale projects. This case analysis contributes to the discussion about multi-method research designs and their role in digital government research. Insights into management strategies specifically designed to respond to the digital government context and the adoption of relevant methodologies drawn from the experiences of the authors are provided.
An analysis framework is presented that can be used to assess important issues of mobile workforce solutions, based on recent literature on mobile business. Within this framework we did two exploratory mobile workforce case studies, one at a rail company, and one at a highway inspection organization, both in the Netherlands. Preliminary evidence shows that mobile workforce solutions have a direct impact on many aspects of an organization’s work practices, though they do not radically change the organizational structure and business processes within a company. The results of the case studies indicate a need for an engineering approach that integrates a consideration of the potential impact of mobile workforce solutions on an organization, its business process and the individuals who work for it. In this study the foundations are laid for further research into improving business processes using mobile workforce solutions.
Decomposition of electromyogram (EMG) provides a valuable means of obtaining motor unit recruitment and firing rate information. The feasibility of decomposing surface EMG signals into their constituent motor unit action potential (MUAP) trains using independent component analysis (ICA) was examined using simulated EMG data. Surface EMG signals detected with an array of nine electrodes were simulated when nine motor units were active. The electrodes were positioned in three different locations with respect to muscle fiber orientation. It was found that ICA based on an instantaneous mixing model was not able to separate all the MUAP trains due to shape variations and time delays between the surface action potentials detected at the different electrode locations. However, from certain independent components, the firing information of a very limited number of motor units could be obtained. This suggests that ICA based on an instantaneous mixing model may yield firing rate information of a small number of motor units from surface EMG signals recorded at relatively high force levels. However, to obtain more information, blind source separation techniques addressing a more complex convolutive mixing model are required. Similar results were obtained for each of three different electrode locations and orientations.
The time varying human multijoint arm dynamics can be modeled by two factors, simplified musculoskeletal dynamics and the uncertainty factor consisting of measurement noises and modeling error of a rigid body dynamics. In some cases, the uncertainty factor may not be Gaussian; the Kalman filter is no longer the optimal filter. In this paper, for the non-Gaussian environment, a recursive filter design method for estimating time varying human multijoint arm viscoelasticity during the arm is moving is presented. The method is based on a score function approach associated with D U factorization algorithm and equivalent noise technique for multiple innovations process. The proposed method for an experimentbased human arm model provides greater accuracy and robustness in capturing texture information of the model under the case of non-Gaussian noises, while the performance of standard Kalman filter degrades significantly.
In this paper we consider the problem of retrieving objects from a database that respond to a pursued goal (relevancy) by the extractors or decision makers to form a short list from which the final selection will be made because it is well known that human beings are good discriminators when facing few objects but perform very poorly in face of a great number of objects. These problems arise in domains such as finance and e-business (selecting a group of firms from a stock exchange database in which to invest, extracting desired goods to buy from a website, etc.), logistics (selecting a group of potential suppliers for a given product from a group of suppliers), administration (selecting a set of potential sites where facilities such as hospital, school, airport ... can be located). Most of the time the objects to be extracted are described by many attributes that can be measured, observed or supplied by experts. The purpose of this paper is to establish a method that permit to extract, from a database, a short list of objects relevant to the pursued goal using information about attributes and decision makers’ preferences. We argue that, for a given goal, there are, almost always, attributes that work towards this goal and those that work against the goal; from this observation we propose to use the satisficing game theory approach that is based on the notion of being good enough as the underlying mathematical tool to establish the extraction method. The main idea of the method that will be established in this paper is to exploit the positive (working towards the pursued goal)/negative (working against the pursued goal) properties of attributes along with decision makers’ preferences expressed by weighting these attributes to construct two measures known as selectability (related to positive attributes) and rejectability (related to negative attributes) in the framework of satisficing game theory. The objects arguable of being good enough are those for which the selectability measure exceeds the rejectability measure in some sense.
A faster and more accurate method for the detection of corneal thickness using an ultrasonic pulse-echo technique was proposed. The corneal thickness was calculated from the time interval between the two echoes reflected at front and rear interfaces of the cornea. The time interval between two echoes was obtained by detecting the peaks of the digitized echoes. A phase-adjusting method, which is superior to the often-used interpolation approach, was used to improve the time resolution to 1/20 of the sampling period. Simulation studies showed this method is comparable to the interpolation method in accuracy at a majority of signal to noise conditions and dynamic ranges of the A/D converter but much easier to implement. The algorithm was implemented in a MCS-8031 microprocessor based pachymeter for measuring corneal thickness. The center frequency of the ultrasonic transducer was 30MHz and the echoes were digitized at a 180MHz sampling frequency. The device showed an accuracy of 1 μm for 10 repeated measurements.
In this paper, we focus on engineering Pareto-optimal digital circuits given the expected input/output behaviour with a minimal design effort. The design objectives to be minimised are: hardware area, response time and power consumption. We do so using the Strength Pareto Evolutionary Algorithms. This is novel application of multi-objective optimisation to circuit design. The performance and quality of the circuits evolved for some benchmarks are presented then compared to those of single objective genetic algorithms as well as to the circuits obtained by human designers. We show that the evolutionary hardware is far better with respect to all objectives than those designed using traditional methods.
Frequently, a set of conserved (common) motifs in a group of functionally related biological sequences (DNA/RNA or proteins) has a specific biological function. Determination of compact groups of these common patterns is a prerequisite for their efficient modeling. We discuss some of the most crucial aspects of computer implementation of three heuristic algorithms which can be used in computational extraction of such conserved motifs from a set of unaligned DNA/RNA sequences. The algorithms included are tabu search, simulated annealing and a population-based genetic algorithm. A server with these algorithms implemented is available as a public web application free for academic and non-profit users at http://sdmc.i2r.a-star.edu.sg/DRAGON/Motif_Search/. This tool can be directly applied in determination of functional patterns in DNA and RNA.
ABSTRACT The use of Group Support Systems has been researched repeatedly using many task-type. For brainstorming tasks the process support of the system have been restricted to ensuring member anonymity and allowing the simultaneous entry of ideas. Little work has been done investigating other process improvements for idea-generating groups. In this paper is investigated the effect on idea quality and quantity of decomposing a brainstorming task with a South African setting. Using Team Expert Choice, an Analytic Hierarchy Process (AHP) based group support system. An experiment was conducted with two groups. It was hypothesized that task decomposition will generate more and better quality ideas. The findings show that task decomposition resulted in 40% more ideas than no decomposition; the effect on decision quality is statistically significant only when decision quality is measured as the number of good ideas. Keywords: Group Support Systems, AHP, Decision making, Brainstorming.
Two different methods to accelerate the search of a Multi-Objective Evolutionary Algorithm (MOEA) using Artificial Neural Networks are presented. Two different methods are proposed. One using ANN to approximate the fitness of the solutions alternated with the real fitness evaluation, being the ANN approximation used only when the estimated error of the neural network was lower than a pre-defined value. In the second method, the ANN is used as a local search strategy by defining new better solutions from the precedent generation. These methods can substantially reduce the number of fitness evaluations on computational expensive problems while not compromise the good search capabilities of MOEA. The efficiency of the methods proposed is tested on several benchmark functions as well on a real multi-optimization problem of polymer extrusion.
While evolutionary computing inspired approaches to multi-objective optimization have many advantages over conventional approaches; they generally do not explicitly exploit directional/gradient information. This can be inefficient if the underlying objectives are reasonably smooth, and this may limit the application of such approaches to real-world problems. This paper develops a local framework for such problems by geometrically analyzing the multiobjective concepts of descent, diversity and convergence/optimality. It is shown that locally optimal, multi-objective descent direction can be calculated that maximally reduce all the objectives and a local sub-space also exists that is a basis for diversity updates. Local convergence of a point towards the optimal Pareto set is therefore assured. The concept of a population of points is also considered and it is shown that it can be used to locally increase the diversity of the population while still ensuring convergence and a method to extract the local directional information from the population is also described. The paper describes and introduces the basic theoretical concepts as well as demonstrating how they are used on simple test problems.
In most large organizations the management of employees with diverse cultures and backgrounds can become complex and sometimes very difficult. Issues such as communication and collaboration within such an environment are normally a problem to deal with. It is believed that technologies such as Intranets, if used correctly, may have a positive impact in this regard. The author of this article investigated the impact of an Intranet on human behaviour in a large insurance company (“The Insurance Company”). The research study shows that Intranets have a very positive impact on communication and collaboration within a multi cultural working environment and as such contribute to the goals set in the IT protocol of the Southern African Development Community (SADC).
