Artificial wavelet neural network and its application in neuro-fuzzy models

In the proposed work, two types of artificial neural networks are proposed by using well-known advantages and valuable features of wavelets and sigmoidal activation functions. Two neurons are derived by adding and multiplying the outputs of the wavelet and the sigmoidal activation functions. These neurons in a feed-forward single hidden layer network result summation wavelet neural network (SWNN) and multiplication wavelet neural network (MWNN). An algorithm is introduced for structure determination of the proposed networks. Approximation properties of SWNN and MWNN have been evaluated with different wavelet functions. The above networks in the consequent part of the neuro-fuzzy model result summation wavelet neuro-fuzzy (SWNF) and multiplication wavelet neuro-fuzzy (MWNF) models. Different types of wavelet function are tested with the proposed networks and fuzzy models on four different dynamical examples. Convergence of the learning process is also guaranteed by adaptive learning rate and performing stability analysis using Lyapunov function.     

Local recurrent sigmoidal-wavelet neurons in feed-forward neural network for forecasting of dynamic systems: Theory

In this paper different structure of the neurons in the hidden layer of a feed-forward network, for forecasting of the dynamic systems, are proposed. Each neuron in the network is a combination of the sigmoidal activation function (SAF) and wavelet activation function (WAF). The output of the hidden neuron is the product of the output from these two activation functions. A delay element is used to feedback the output of the sigmoidal and the wavelet activation function to each other. This arrangement leads to proposed five possible configurations of recurrent neurons. Besides proposing these neuron models, the presented paper tries to compare the performance of wavelet function with sigmoid function. To guarantee the stability and the convergence of the learning process, upper bound for the learning rates has been investigated using the Lyapunov stability theorem. A two-phase adaptive learning rate ensures this upper bound. Universal approximation property of the feed-forward network with the proposed neurons has also been investigated. Finally, the applicability and comparison of the proposed recurrent networks has been weathered on two benchmark problem catering different types of dynamical systems.     

Structure identification of generalized adaptive neuro-fuzzy inference systems

This paper presents a method to identify the structure of generalized adaptive neuro-fuzzy inference systems (GANFISs). The structure of GANFIS consists of a number of generalized radial basis function (GRBF) units. The radial basis functions are irregularly distributed in the form of hyper-patches in the input-output space. The minimum number of GRBF units is selected based on a heuristic using the fuzzy curve. For structure identification, a new criterion called structure identification criterion (SIC) is proposed. SIC deals with a trade off between performance and computational complexity of the GANFIS model. The computational complexity of gradient descent learning is formulated based on simulation study. Three methods of initialization of GANFIS, viz., fuzzy curve, fuzzy C-means in x/spl times/y space and modified mountain clustering have been compared in terms of cluster validity measure, Akaike's information criterion (AIC) and the proposed SIC.     

Blood flow and coherent vortices in the normal and aneurysmatic aortas: a fluid dynamical approach to intra-luminal thrombus formation

Abdominal aortic aneurysms (AAAs) are frequently characterized by the development of an intra-luminal thrombus (ILT), which is known to have multiple biochemical and biomechanical implications. Development of the ILT is not well understood, and shear–stress-triggered activation of platelets could be the first step in its evolution. Vortical structures (VSs) in the flow affect platelet dynamics, which motivated the present study of a possible correlation between VS and ILT formation in AAAs. VSs educed by the λ₂-method using computational fluid dynamics simulations of the backward-facing step problem, normal aorta, fusiform AAA and saccular AAA were investigated. Patient-specific luminal geometries were reconstructed from computed tomography scans, and Newtonian and Carreau–Yasuda models were used to capture salient rheological features of blood flow. Particularly in complex flow domains, results depended on the constitutive model. VSs developed all along the normal aorta, showing that a clear correlation between VSs and high wall shear stress (WSS) existed, and that VSs started to break up during late systole. In contrast, in the fusiform AAA, large VSs developed at sites of tortuous geometry and high WSS, occupying the entire lumen, and lasting over the entire cardiac cycle. Downward motion of VSs in the AAA was in the range of a few centimetres per cardiac cycle, and with a VS burst at that location, the release (from VSs) of shear-stress-activated platelets and their deposition to the wall was within the lower part of the diseased artery, i.e. where the thickest ILT layer is typically observed. In the saccular AAA, only one VS was found near the healthy portion of the aorta, while in the aneurysmatic bulge, no VSs occurred. We present a fluid-dynamics-motivated mechanism for platelet activation, convection and deposition in AAAs that has the potential of improving our current understanding of the pathophysiology of fluid-driven ILT growth.     

Approximate Analytical Solution of Stagnation Point Flow and Heat Transfer over an Exponential Stretching Sheet with Convective Boundary Condition

This study is concerned with the stagnation point flow and heat transfer over an exponential stretching sheet via an approximate analytical method known as optimal homotopy asymptotic method (OHAM). The governing partial differential equations are converted into ordinary nonlinear differential equations using similarity transformations available in the literature. The heat transfer problem is modeled using two‐point convective boundary condition. These equations are then solved using the OHAM approach. The effects of controlling parameters on the dimensionless velocity, temperature, friction factor, and heat transfer rate are analyzed and discussed through graphs and tables. It is found that the OHAM results match well with numerical results obtained by Runge–Kutta Fehlberg fourth‐fifth order method for different assigned values of parameters. The rate of heat transfer increases with the stretching parameter. It is also found that the stretching parameter reduces the hydrodynamic boundary layer thickness whereas the Prandtl number reduces the thermal boundary layer thickness.     

Bed-Form Geometry in Sand-Bed Flows

A new method is proposed for predicting relative bed-form height h∕d in sand-bed flows. The proposed method is based on the concept of relating energy loss due to form drag to the head loss across a sudden expansion in open channel flows. A unique feature of the proposed method is that it can be applied to various bed forms, i.e., ripples, dunes, antidunes∕standing waves, and transitional bed regimes that occur in alluvial flows. The relation thus developed was applied to a large number (251 flows, 14 different data sets) of laboratory and river data, and was found to give good agreement with the observed h∕d values. In a comparison of prediction accuracies with seven existing relationships, the proposed method was found to give significantly better agreements with the observed data. Future improvements in the prediction of h∕d will depend on improved formulations of the two parameters incorporated in the present relation, i.e., energy loss coefficient K, and the relative bed-form length L∕d for various bed configurations. More research is needed to develop better formulations for these parameters.     

Generating Adversarial Samples on Multivariate Time Series using Variational Autoencoders

Classification models for multivariate time series have drawn the interest of many researchers to the field with the objective of developing accurate and efficient models. However, limited research has been conducted on generating adversarial samples for multivariate time series classification models. Adversarial samples could become a security concern in systems with complex sets of sensors. This study proposes extending the existing gradient adversarial transformation network (GATN) in combination with adversarial autoencoders to attack multivariate time series classification models. The proposed model attacks classification models by utilizing a distilled model to imitate the output of the multivariate time series classification model. In addition, the adversarial generator function is replaced with a variational autoencoder to enhance the adversarial samples. The developed methodology is tested on two multivariate time series classification models: 1-nearest neighbor dynamic time warping (1-NN DTW) and a fully convolutional network (FCN). This study utilizes 30 multivariate time series benchmarks provided by the University of East Anglia (UEA) and University of California Riverside (UCR). The use of adversarial autoencoders shows an increase in the fraction of successful adversaries generated on multivariate time series. To the best of our knowledge, this is the first study to explore adversarial attacks on multivariate time series. Additionally, we recommend future research utilizing the generated latent space from the variational autoencoders.      

A nonlinear model for optimizing the performance of a multi‐product production line

This paper examines the measures of performance and in particular addresses the throughput of an automated production line processing multiple products. The line is composed of a sequence of workstations connected in series with finite buffers in between. We explore the effects of buffer size on attenuating the impact of line blocking and starvation that can cause a reduction in the output. Such effects are analyzed through a nonlinear mathematical programming model and the implications are examined. The aim of the model is to achieve the best performance subject to available workstation capacity without overexpenditure on buffer size. Single and multi‐objective optimizations are carried out in the paper. A numerical example of a production line with a given configuration of workstations; workstation capacity; and job mix is presented to demonstrate the model and its application. A discussion of the impact of buffer size on maximum throughput is also provided. The paper is concluded with a discussion on the decision‐making implications.