Neural Network Based Feedback Linearization Control of an Unmanned Aerial Vehicle

This paper presents a flight control design for an unmanned aerial vehicle （UAV） using a nonlinear autoregressive moving average （NARMA-L2） neural network based feedback linearization and output redefinition technique. The UAV investigated is non- minimum phase. The output redefinition technique is used in such a way that the resulting system to be inverted is a minimum phase system. The NARMA-L2 neural network is trained off-line for forward dynamics of the UAV model with redefined output and is then inverted to force the real output to approximately track a command input. Simulation results show that the proposed approaches have good performance.

Global stability of interval recurrent neural networks

The robust global exponential stability of a class of interval recurrent neural networks（RNNs） is studied,and a new robust stability criterion is obtained in the form of linear matrix inequality.The problem of robust stability of interval RNNs is transformed into a problem of solving a class of linear matrix inequalities.Thus,the robust stability of interval RNNs can be analyzed by directly using the linear matrix inequalities（LMI） toolbox of MATLAB.Numerical example is given to show the effectiveness of the obtained results.

Wind Power Forecasting Using Wavelet Transforms and Neural Networks with Tapped Delay

With an objective to improve wind power estimation accuracy and reliability,this paper presents Linear Neural Networks with Tapped Delay(LNNTD)in combination with wavelet transform(WT)for probabilistic wind power forecasting in a time series framework.For comparison purposes,results of the proposed model are compared with the benchmark model,different neural networks and WT based models considering performance indices such as accuracy,execution time and R^(2) statistic.For the reliability and proper validation of the proposed model,this paper highlights the probabilistic forecast attributes at different skill tests.The historical data of the Ontario Electricity Market(OEM)for the period 2011–2014 were used and tested for two years from November 2012 to October 2014 with one month moving window considering all seasonal aspects.The experimental results clearly show that the results of the proposed model have been found to be better than others.

Information criterion based fast PCA adaptive algorithm

The novel information criterion （NIC） algorithm can find the principal subspace quickly, but it is not an actual principal component analysis （PCA） algorithm and hence it cannot find the orthonormal eigen-space which corresponds to the principal component of input vector. This defect limits its application in practice. By weighting the neural network＇s output of NIC, a modified novel information criterion （MNIC） algorithm is presented. MNIC extractes the principal components and corresponding eigenvectors in a parallel online learning program, and overcomes the NIC＇s defect. It is proved to have a single global optimum and nonquadratic convergence rate, which is superior to the conventional PCA online algorithms such as Oja and LMSER. The relationship among Oja, LMSER and MNIC is exhibited. Simulations show that MNIC could converge to the optimum fast. The validity of MNIC is proved.

Verifying ReLU Neural Networks from a Model Checking Perspective

Neural networks, as an important computing model, have a wide application in artificial intelligence (AI) domain. From the perspective of computer science, such a computing model requires a formal description of its behaviors, particularly the relation between input and output. In addition, such specifications ought to be verified automatically. ReLU (rectified linear unit) neural networks are intensively used in practice. In this paper, we present ReLU Temporal Logic (ReTL), whose semantics is defined with respect to ReLU neural networks, which could specify value-related properties about the network. We show that the model checking algorithm for theΣ2∪Π2 fragment of ReTL, which can express properties such as output reachability, is decidable in EXPSPACE. We have also implemented our algorithm with a prototype tool, and experimental results demonstrate the feasibility of the presented model checking approach.

An application of local linear radial basis function neural network for flood prediction

Heavy seasonal rain makes waterway flood and is one of the preeminent reason behind flooding.Flooding causes various perils with outcomes including danger to human life,harm to building,streets,misfortune to horticultural fields and bringing about human uprooting.Thus,prediction of flood is of prime importance so as to reduce exposure of people and destruction of property.This paper focuses on applying different neural networks approach,i.e.Multilayer Perceptron,Radial Basis functional neural network,Local Linear Radial Basis Functional Neural Network and Artificial Neural Network with Whale Optimization to predict flood in terms of rainfall,gauge,area,velocity,pressure,average temperature,average wind speed that are setup through field and lab investigation from the contextual analysis of river“Daya”and“Bhargavi”.It has always been a troublesome undertaking to predict flood as many factors have influence on it although with this neural network models the prediction accuracy can be optimized using back propagation method which is a widely applied over traditional learning method for neural system because of its preeminent learning ability.The flood prediction system is built with the four models and a comparison is made which provides us the answer to which model is effective for the prediction.

An IoT Based Secure Patient Health Monitoring System

Internet of things(IoT)field has emerged due to the rapid growth of artificial intelligence and communication technologies.The use of IoT technology in modern healthcare environments is convenient for doctors and patients as it can be used in real-time monitoring of patients,proper administration of patient information,and healthcare management.However,the usage of IoT in the healthcare domain will become a nightmare if patient information is not securely maintainedwhile transferring over an insecure network or storing at the administrator end.In this manuscript,the authors have developed a secure IoT healthcare monitoring system using the Blockchainbased XOR Elliptic Curve Cryptography(BC-XORECC)technique to avoid various vulnerable attacks.Initially,thework has established an authentication process for patient details by generating tokens,keys,and tags using Length Ceaser Cipher-based PearsonHashingAlgorithm(LCC-PHA),EllipticCurve Cryptography(ECC),and Fishers Yates Shuffled Based Adelson-Velskii and Landis(FYS-AVL)tree.The authentications prevent unauthorized users from accessing or misuse the data.After that,a secure data transfer is performed using BC-XORECC,which acts faster by maintaining high data privacy and blocking the path for the attackers.Finally,the Linear Spline Kernel-Based Recurrent Neural Network(LSK-RNN)classification monitors the patient’s health status.The whole developed framework brings out a secure data transfer without data loss or data breaches and remains efficient for health care monitoring via IoT.Experimental analysis shows that the proposed framework achieves a faster encryption and decryption time,classifies the patient’s health status with an accuracy of 89%,and remains robust comparedwith the existing state-of-the-art method.

Improving time series forecasting using elephant herd optimization with feature selection methods

The time series data is chaotic,non seasonal,non stationary and random in nature.It becomes quite challenging to discover the hidden patterns of time series data.In this paper the time series data is predicted with the help of a machine learning algorithm i.e.Elephant Herd Optimization(EHO).Three different types of time series data are used to testify the superiority of the proposed method namely stock market data,currency exchange data and absenteeism at work.The data are first subjected to feature selection methods namely ANOVA and Friedman test.The feature selection methods provide relevant set of features which is fed to the neural network trained with the method.The proposed method is also compared with other methods such as local linear radial basis functional neural network and particle swarm optimization.The results prove supremacy of EHO over other methods.

Fast adaptive principal component extraction based on a generalized energy function

By introducing an arbitrary diagonal matrix, a generalized energy function (GEF) is proposed for searching for the optimum weights of a two layer linear neural network. From the GEF, we derive a recur- sive least squares (RLS) algorithm to extract in parallel multiple principal components of the input covari- ance matrix without designing an asymmetrical circuit. The local stability of the GEF algorithm at the equilibrium is analytically verified. Simulation results show that the GEF algorithm for parallel multiple principal components extraction exhibits the fast convergence and has the improved robustness resis- tance to the eigenvalue spread of the input covariance matrix as compared to the well-known lateral inhi- bition model (APEX) and least mean square error reconstruction (LMSER) algorithms.