EEG signals classification based on autoregressive and inherently quantum recurrent neural network

This paper presents a new blind direction of arrival (DOA) estimation approach for closely-spaced sources. The new method first estimates the autoregressive (AR) coefficients via an initial DOA estimation and then uses the AR coefficients for the linear extrapolation of the correlation matrix to implement a fine DOA estimation. Both initial and fine DOA estimations are performed using the estimation of signal parameters via rotational invariance techniques (ESPRIT) algorithm. Unlike a conventional AR coefficient estimation method which estimates the AR coefficients on the snapshot basis, our AR coefficients are estimated in the correlation domain once for a block of snapshots, thus significantly reducing the computational complexity of the antenna array. Moreover, the proposed spatial extrapolation-based DOA estimation approach is analyzed using perturbation theory. Both the theoretical analysis and computer simulations show that the proposed method outperforms the conventional techniques in terms of the mean square error (MSE) of the DOA estimation when the angle of separation of the signal sources is very small.

Classification of electroencephalogram (EEG) signals is an important task in the brain computer interface system. This paper presents two combination strategies of feature extraction on EEG signals. In the first strategy, Autoregressive coefficients and approximate entropy are calculated respectively, and the features are obtained by assembling them. In the second strategy, the EEG signals are first decomposed into sub-bands by wavelet packet decomposition. Wavelet packet coefficients are then sent to the autoregressive model to calculate autoregressive coefficients, which are used as features extracted from the original EEG signals. These features are fed to support vector machine for classifying the EEG signals. The classification accuracy has been used for evaluating the classification performance. Experimental results in five mental tasks show that the combination strategies can effectively improve the classification performance when the order of autoregressive model is greater than 5, and the second strategy is superior to the first one in terms of the classification accuracy.

Quantum Neural Networks (QNN) is a burgeoning new field built upon the combination of classical neural networks and quantum computations, which has many problems needed to solve. The predictive model of QNN is an issue that must be settled to develop QNN based on proportional integral derivative neural networks and quantum computation, which can be so called generalized quantum neural networks (GQNN). Firstly, we describe the theory of quantum computation and neural networks. Secondly, it can realize the algorithm of prediction to construct the modeling of generalized quantum neural networks for those complexity nonlinear systems. Finally, using an example explains the model of generalized quantum neural networks. The computational results shows that GQNN is more effective than conventional neural networks.

The Sixth International Symposium on Neural Networks (ISNN 2009)

According to recent research on the brain and cognition, the microtubule level activities in the brain are in accordance with the quantum mechanical concepts. Consciousness is the emergent phenomenon of the brain’s subsystems and the quantum neural correlates. Based on the global work-space theory and traditional neural networks, investigations in machine consciousness and machine intelligence are reporting new techniques. In this study, a novel approach using circuit-based quantum neural network is proposed and simulated. This approach satisfies all the criteria of quantum computing and is tested for the exclusive OR (XOR) gate’s nonlinear learning. As a result of the use of quantum gates, various quantum circuits, such as quantum adders and subtractors, are also created and included in the designing and simulation of circuit of the quantum neural networks. Moreover, it is also argued that the proposed circuit of quantum neural network may also be tested and implemented on real quantum computer hardware. The present study also stresses the applicability of techniques of machine learning algorithms such as quantum and classical neural networks to various challenges of High Energy Physics.

Adding data process feedback to the nonlinear autoregressive model

Investigation of the diagnosis of human brain through the electroencephalograph (EEG) is an important application of EEG signals. While automated techniques exist for EEG analysis, it is likely that additional information can be extracted from EEG signals through the use of new methods. In this paper, we propose a method that applies artificial neural networks approach for identification of the inert region from EEG signals. The method has three main steps. First, a large scale, complex EEG signals is simply normalized into a reasonable data. Second, characteristic feature from EEG signals, the normalized EEG data has no regular rule and is still so complex that neural network can not learn the characteristic feature from EEG. Hence we extract characteristic feature from EEG and use principal component analysis (PCA) and root mean square (RMS) processes in EEG data. Finally, neural network is supplied as the input value to these processes in EEG and learn these characteristic feature. To demonstrate the effectiveness of the method, we perform simulations on location of inert region from EEG data, consists of 10000 training and test data. These EEG estimation tasks were created by using a set of calculated, artificial EEG signals based on a number of current dipoles. The experimental results indicate that the proposed method has several attractive features. 1) The method can estimate the EEG feature of inert region and better generalization performance can be achieved than non preprocessing EEG. 2) The more larger inert region were, the more small estimation error become.

In the previous chapters, we discussed autoregressive linear mixed effects models. In this section, we discuss the relationships between the autoregressive linear mixed effects models and nonlinear mixed effects models, growth curves, and differential equations. The autoregressive model shows a profile approaching an asymptote, where the change is proportional to the distance remaining to the asymptote. Autoregressive models in discrete time correspond to monomolecular curves in continuous time. Autoregressive linear mixed effects models correspond to monomolecular curves with random effects in the baseline and asymptote, and special error terms. The autoregressive coefficient is a nonlinear parameter, but all random effects parameters in the model are linear. Therefore, autoregressive linear mixed effects models are nonlinear mixed effects models without nonlinear random effects and have a closed form of likelihood. When there are time-dependent covariates, autoregressive linear mixed effects models are represented by a differential equation and random effects. The monomolecular curve is one of the popular growth curves. We introduce other growth curves, such as the logistic curves and von Bertalanffy curves, and generalizations of growth curves. Re-parameterization is often performed in nonlinear models, and various representations of re-parameterization in monomolecular and other curves are provided herein.

Nowadays , nonlinear time series and artificial neural networks (ANN) models are used for forecasting in the field of business, agriculture and soon. Recent studies have shown, ANN have been successfully used for forecasting of financial and agriculture data series The classical methods used for time series prediction like Box-Jenkins or ARIMA assumes that there is a linear relationship between inputs and outputs. ANN have more advantages that can approximate to model both linear and nonlinear structures in time series, they are not able to handling both structures equally well. The autoregressive integrated moving average (ARIMA) model and two ANN models namely, Radial basis function neural networks (RBFNN), and Elman recurrent neural networks (ERNN) methods were applied to Hyderabad airport traffic data. The data obtained for 15 years from 2002–2003 to 2016–2017 about domestic and international passenger of International Airport of Hyderabad, India. In this research paper, we compared the performances of ARIMA, RBFNN and ERNN were based on three measures: mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE). The results showed that RBFNN obtained the smallest MAE, MAPE and RMSE in both the modeling and forecasting processes. The performances of the three models ranked in ascending order were: ARIMA, ERNN and the RBFNN model. Keywords: T ime series, forecasting, artificial neural networks, ARIMA models, radial basis function neural networks, and Elman recurrent neural networks Cite this Article R. Ramakrishna, Berhe Aregay, Tewodros Gebregergs. The Comparison in Time Series Forecasting of Air Traffic Data by Autoregressive Integrated Moving Average Model, Radial Basis Function and Elman Recurrent Neural Networks. Research & Reviews: Journal of Statistics . 2018; 7(3): 75–90p.

With the recent growth of the Internet of Things (IoT) and the demand for faster computation, quantized neural networks (QNNs) or QNN-enabled IoT can offer better performance than conventional convolution neural networks (CNNs). With the aim of reducing memory access costs and increasing the computation efficiency, QNN-enabled devices are expected to transform numerous industrial applications with lower processing latency and power consumption. Another form of QNN is the binarized neural network (BNN), which has 2 bits of quantized levels. In this paper, CNN-, QNN-, and BNN-based pattern recognition techniques are implemented and analyzed on an FPGA. The FPGA hardware acts as an IoT device due to connectivity with the cloud, and QNN and BNN are considered to offer better performance in terms of low power and low resource use on hardware platforms. The CNN and QNN implementation and their comparative analysis are analyzed based on their accuracy, weight bit error, RoC curve, and execution speed. The paper also discusses various approaches that can be deployed for optimizing various CNN and QNN models with additionally available tools. The work is performed on the Xilinx Zynq 7020 series Pynq Z2 board, which serves as our FPGA-based low-power IoT device. The MNIST and CIFAR-10 databases are considered for simulation and experimentation. The work shows that the accuracy is 95.5% and 79.22% for the MNIST and CIFAR-10 databases, respectively, for full precision (32-bit), and the execution time is 5.8 ms and 18 ms for the MNIST and CIFAR-10 databases, respectively, for full precision (32-bit).

Detection of alertness-related EEG signals based on decision fused BP neural network

Deep neural networks (DNNs) are state-of-the-art algorithms for multiple applications, spanning from image classification to speech recognition. While providing excellent accuracy, they often have enormous compute and memory requirements. As a result of this, quantized neural networks (QNNs) are increasingly being adopted and deployed especially on embedded devices, thanks to their high accuracy, but also since they have significantly lower compute and memory requirements compared to their floating point equivalents. QNN deployment is also being evaluated for safety-critical applications, such as automotive, avionics, medical or industrial. These systems require functional safety, guaranteeing failure-free behaviour even in the presence of hardware faults. In general fault tolerance can be achieved by adding redundancy to the system, which further exacerbates the overall computational demands and makes it difficult to meet the power and performance requirements. In order to decrease the hardware cost for achieving functional safety, it is vital to explore domain-specific solutions which can exploit the inherent features of DNNs. In this work we present a novel methodology called fault-aware training (FAT), which includes error modeling during neural network (NN) training, to make QNNs resilient to specific fault models on the device. Our experiments show that by injecting faults in the convolutional layers during training, highly accurate convolutional neural networks (CNNs) can be trained which exhibits much better error tolerance compared to the original. Furthermore, we show that redundant systems which are built from QNNs trained with FAT achieve higher worse-case accuracy at lower hardware cost. This has been validated for numerous classification tasks including CIFAR10, GTSRB, SVHN and ImageNet.

BackgroundReliable evaluations of state-level policies are essential for identifying effective policies and informing policymakers’ decisions. State-level policy evaluations commonly use a difference-in-differences (DID) study design; yet within this framework, statistical model specification varies notably across studies. More guidance is needed about which set of statistical models perform best when estimating how state-level policies affect outcomes.MethodsMotivated by applied state-level opioid policy evaluations, we implemented an extensive simulation study to compare the statistical performance of multiple variations of the two-way fixed effect models traditionally used for DID under a range of simulation conditions. We also explored the performance of autoregressive (AR) and GEE models. We simulated policy effects on annual state-level opioid mortality rates and assessed statistical performance using various metrics, including directional bias, magnitude bias, and root mean squared error. We also reported Type I error rates and the rate of correctly rejecting the null hypothesis (e.g., power), given the prevalence of frequentist null hypothesis significance testing in the applied literature.ResultsMost linear models resulted in minimal bias. However, non-linear models and population-weighted versions of classic linear two-way fixed effect and linear GEE models yielded considerable bias (60 to 160%). Further, root mean square error was minimized by linear AR models when we examined crude mortality rates and by negative binomial models when we examined raw death counts. In the context of frequentist hypothesis testing, many models yielded high Type I error rates and very low rates of correctly rejecting the null hypothesis (< 10%), raising concerns of spurious conclusions about policy effectiveness in the opioid literature. When considering performance across models, the linear AR models were optimal in terms of directional bias, root mean squared error, Type I error, and correct rejection rates.ConclusionsThe findings highlight notable limitations of commonly used statistical models for DID designs, which are widely used in opioid policy studies and in state policy evaluations more broadly. In contrast, the optimal model we identified--the AR model--is rarely used in state policy evaluation. We urge applied researchers to move beyond the classic DID paradigm and adopt use of AR models.

This paper presents a classification method for arrhythmia using probabilistic neural network, based on unique combination of three Electrocardiogram (ECG) features; heart rate, Auto Regressive (AR) coefficients & spectral entropy (SE). Heart rate has been very critical parameter for the detection of life-threatening arrhythmia. The purpose of this paper is to develop a Probabilistic Neural Network (PNN) based algorithm for improved detection and broader classification of cardiac arrhythmia. The results show that the unique combination of ECG features considered in this work provides more accurate and robust classification of arrhythmias.

Quantum neural networks (QNNs) represent an emerging technology that uses a quantum computer for neural network computations. The QNNs have demonstrated potential advantages over classical neural networks in certain tasks. As a core component of a QNN, the parameterized quantum circuit (PQC) plays a crucial role in determining the QNN's overall performance. However, quantum circuit architectures designed manually based on experience or using specific hardware structures can suffer from inefficiency due to the introduction of redundant quantum gates, which amplifies the impact of noise on system performance. Recent studies have suggested that the advantages of quantum evolutionary algorithms (QEAs) in terms of precision and convergence speed can provide an effective solution to quantum circuit architecture-related problems. Currently, most QEAs adopt a fixed rotation mode in the evolution process, and a lack of an adaptive updating mode can cause the QEAs to fall into a local optimum and make it difficult for them to converge. To address these problems, this study proposes an adaptive quantum evolution algorithm (AQEA). First, an adaptive mechanism is introduced to the evolution process, and the strategy of combining two dynamic rotation angles is adopted. Second, to prevent the fluctuations of the population's offspring, the elite retention of the parents is used to ensure the inheritance of good genes. Finally, when the population falls into a local optimum, a quantum catastrophe mechanism is employed to break the current population state. The experimental results show that compared with the QNN structure based on manual design and QEA search, the proposed AQEA can reduce the number of network parameters by up to 20% and increase the accuracy by 7.21%. Moreover, in noisy environments, the AQEA-optimized circuit outperforms traditional circuits in maintaining high fidelity, and its excellent noise resistance provides strong support for the reliability of quantum computing.