Knowledge extraction from a mixed transfer function artificial neural network

One of the main problems with Artificial Neural Networks (ANNs) is that their results are not intuitively clear. For example, commonly used hidden neurons with sigmoid activation function can approximate any continuous function, including linear functions, but the coefficients (weights) of this approximation are rather meaningless. To address this problem, current paper presents a novel kind of a neural network that uses transfer functions of various complexities in contrast to mono-transfer functions used in sigmoid and hyperbolic tangent networks. The presence of transfer functions of various complexities in a Mixed Transfer Functions Artificial Neural Network (MTFANN) allow easy conversion of the full model into user-friendly equation format (similar to that of linear regression) without any pruning or simplification of the model. At the same time, MTFANN maintains similar generalization ability to mono-transfer function networks in a global optimization context. The performance and knowledge extraction of MTFANN were evaluated on a realistic simulation of the Puma 560 robot arm and compared to sigmoid, hyperbolic tangent, linear and sinusoidal networks.

Mixed transfer function neural networks for knowledge acquistition

Modeling helps to understand and predict the outcome of complex systems. Inductive modeling methodologies are beneficial for modeling the systems where the uncertainties involved in the system do not permit to obtain an accurate physical model. However inductive models, like artificial neural networks (ANNs), may suffer from a few drawbacks involving over-fitting and the difficulty to easily understand the model itself. This can result in user reluctance to accept the model or even complete rejection of the modeling results. Thus, it becomes highly desirable to make such inductive models more comprehensible and to automatically determine the model complexity to avoid over-fitting. In this paper, we propose a novel type of ANN, a mixed transfer function artificial neural network (MTFANN), which aims to improve the complexity fitting and comprehensibility of the most popular type of ANN (MLP - a Multilayer Perceptron).

Knowledge extraction from a mixed transfer function artificial neural network

One of the big problems with Artificial Neural Networks (ANN) is that their results are not intuitively clear. For example, if we use the traditional neurons, with a sigmoid activation function, we can approximate any function, including linear functions, but the coefficients (weights) in this approximation will be rather meaningless. To resolve this problem, this paper presents a novel kind of ANN with different transfer functions mixed together. The aim of such a network is to i) obtain a better generalization than current networks ii) to obtain knowledge from the networks without a sophisticated knowledge extraction algorithm iii) to increase the understanding and acceptance of ANNs. Transfer Complexity Ratio is defined to make a sense of the weights associated with the network. The paper begins with a review of the knowledge extraction from ANNs and then presents a Mixed Transfer Function Artificial Neural Network (MTFANN). A MTFANN contains different transfer functions mixed together rather than mono-transfer functions. This mixed presence has helped to obtain high level knowledge and similar generalization comparatively to monotransfer function nets in a global optimization context.

Mixed transfer function neural networks for generalization and knowledge extraction

This thesis develops a novel framework of nonlinear modelling to adaptively fit the complexity of the model to the problem domain resulting in a better modelling capability and a straightforward knowledge acquisition. The developed framework also permits increased comprehensibility and user acceptability of modelling results.

The use of prolog in the realisation of digital transfer functions

Systems characterised by general graphs are very flexible because of a given application they offer a great number of degrees of freedom and consequently a great possibility of choice to the designer. This can become a drawback in practice when myriads of cases have to be examined analytically and most of them rejected for a variety of heterogeneous reasons. For such design problems, it appears that logic programming languages are ideally suited because they can implement the heterogeneous rules which describe the desired system in a natural way. An application of the Prolog language to the design of active RC biquad circuits investigated by Mouly in his doctoral thesis illustrates the advantages of such a rule-based approach.

Forecasting inflation, industrial output and exchange rates : a template study for India

This paper reports results from a forecasting study for inflation, industrial output and exchange rates for India. We cannot reject the null hypothesis for linearity for all series used except for the growth rate of the foreign exchange series and our analysis is based on linear models, ARIMA and bivariate transfer functions and restricted VAR. Forecasting performance is evaluated using the models’ root mean-squared error differences and Theil’s inequality coefficients from recursive origin static, fixed origin dynamic and rolling origin dynamic forecasts. For models based on weekly data, based on RMSEs, we find that the bivariate models improve upon the forecasts of the ARIMA model while for models based on monthly data the ARIMA model has almost always better performance. In choosing between the two bivariate models on the basis of RMSEs, our overall results tend to support the use of a restricted VAR, as this model had the best forecasting performance more frequently than the transfer function model.

Functional fabrics with controlled wettability

In this study, fabrics with novel super water-repellent and unidirectional water-transfer functions have been prepared using one-step wet-chemical coating processes. The mechanism of directional water transport across the fabrics having gradient superhydrophobicity to hydrophilicity has been elucidated.

Use of artificial neural networks as explicit finite difference operators

Results of a numerical exercise, substituting a numerical operator by an artificial neural network (ANN) are presented in this paper. The numerical operator used is the explicit form of the finite difference (FD) scheme. The FD scheme was used to discretize the one-dimensional transport equation, which included both the advection and dispersion terms. Inputs to the ANN are the FD representation of the transport equation, and the concentration was designated as the output. Concentration values used for training the ANN were obtained from analytical solutions. The numerical operator was reconstructed from a back calculation of the weights of the ANN. Linear transfer functions were used for this purpose. The ANN was able to accurately recover the velocity used in the training data, but not the dispersion coefficient. This capability was improved when numerical dispersion was taken into account; however, it is limited to the condition: C/P<0.5 , where C is the Courant number and P , the Peclet number (i.e., the restriction imposed by the Neumann stability condition).

Artificial neural network modeling of flow stress in hot rolling

In this study, an artificial neural network model is proposed to predict the flow stress variations during the hot rolling process. Optimization of the proposed neural network with respect to number of neurons within the hidden layer, different training methods and transfer functions of the neural network is performed. The results of the optimal network were compared with those of the conventional analytic method and it is shown that using an optimal neural network the mean calculated error is drastically reduced.

Exponential stabilization of neural networks with various activation functions and mixed time-varying delays

This paper presents some results on the global exponential stabilization for neural networks with various activation functions and time-varying continuously distributed delays. Based on augmented time-varying Lyapunov-Krasovskii functionals, new delay-dependent conditions for the global exponential stabilization are obtained in terms of linear matrix inequalities. A numerical example is given to illustrate the feasibility of our results.

A secure tag ownership transfer scheme in a closed loop RFID system

In this paper we propose a novel secure tag ownership transfer scheme for closed loop RFID systems. An important property of our method is that the ownership transfer is guaranteed to be atomic and the scheme is protected against desynchronisation leading to permanent DoS. Further, it is suited to the computational constraints of EPC Class-1 Gen-2 passive RFID tags as they only use the CRC and PRNG functions that passive RFID tags are capable of. We provide a detailed security analysis to show that our scheme satisfies the required security properties of tag anonymity, tag location privacy, forward secrecy, forward untraceability while being resistant to replay, desynchronisation and server impersonation attacks. Performance comparisons show that our scheme is practical and can be implemented on passive low-cost RFID tags.

Function-based approach to mixed haptic effects rendering

Commonly, surface and solid haptic effects are defined in such a way that they hardly can be rendered together. We propose a method for defining mixed haptic effects including surface, solid, and force fields. These haptic effects can be applied to virtual scenes containing various objects, including polygon meshes, point clouds, impostors, and layered textures, voxel models as well as function-based shapes. Accordingly, we propose a way how to identify location of the haptic tool in such virtual scenes as well as consistently and seamlessly determine haptic effects when the haptic tool moves in the scenes with objects having different sizes, locations, and mutual penetrations. To provide for an efficient and flexible rendering of haptic effects, we propose to concurrently use explicit, implicit and parametric functions, and algorithmic procedures.

A potential-controlled switch on/off mechanism for selective excitation in mixed electrochemiluminescent systems

We demonstrate the complete, rapid, and reversible switching between the emissions from two electrogenerated chemiluminescence (ECL) systems contained within the same solution, controlled by simple modification of the applied potential. The fundamental bases of the approach are the ability to selectively 'switch on' luminophores at distinct oxidation potentials, and an intriguing observation that the emission from the well-known electrochemiluminescent complex, fac-Ir(ppy)3, (where ppy is 2-phenylpyridinato), can be selectively 'switched-off' at high overpotentials. The dependence of this phenomenon on high concentrations of the co-reactant implicates quenching of the excited [Ir(ppy)3]* state by electron transfer. Rapid spectral scanning during modulation of the applied potential reveals well resolved maxima for mixtures comprising either green and red or green and blue luminophores, illustrating the vast potential of this approach for multiplexed ECL detection.

Annihilation electrogenerated chemiluminescence of mixed metal chelates in solution : Modulating emission colour by manipulating the energetics

We demonstrate the mixed annihilation electrogenerated chemiluminescence of tris(2,2′-bipyridine)ruthenium(ii) with various cyclometalated iridium(iii) chelates. Compared to mixed ECL systems comprising organic luminophores, the absence of T-route pathways enables effective predictions of the observed ECL based on simple estimations of the exergonicity of the reactions leading to excited state production. Moreover, the multiple, closely spaced reductions and oxidations of the metal chelates provide the ability to finely tune the energetics and therefore the observed emission colour. Distinct emissions from multiple luminophores in the same solution are observed in numerous systems. The relative intensity of these emissions and the overall emission colour are dependent on the particular oxidized and reduced species selected by the applied electrochemical potentials. Finally, these studies offer insights into the importance of electronic factors in the question of whether the reduced or oxidized partner becomes excited in annihilation ECL. This journal is