Feature selection is essential in various fields of science and engineering, from remote sensing to computer vision. Reducing data dimensionality by removing redundant features and selecting the most ...informative ones improves machine learning algorithms' performance, especially in supervised classification tasks, while lowering storage needs. Graph-embedding techniques have recently been found efficient for feature selection since they preserve the geometric structure of the original feature space while embedding data into a low-dimensional subspace. However, the main drawback is the high computational cost of solving an eigenvalue decomposition problem, especially for large-scale problems. This paper addresses this issue by combining the graph embedding framework and representation theory for a novel feature selection method. Inspired by the high dimensional model representation, the feature transformation is assumed to be a linear combination of a set of univariate orthogonal functions carried out in the graph embedding framework. As a result, an explicit embedding function is created, which can be utilised to embed out-of-samples into low-dimensional space and provide a feature relevance score. The significant contribution of the proposed method is to divide an n -dimensional generalised eigenvalue problem into n small-sized eigenvalue problems. With this property, the computational complexity of the graph embedding is significantly reduced, resulting in a scalable feature selection method, which could be easily parallelized too. The performance of the proposed method is compared favourably to its counterparts in high-dimensional hyperspectral image processing in terms of classification accuracy, feature stability, and computational time.
There are emerging trends to use the Industrial Internet of Things (IIoT) in manufacturing and related industries. Machine Learning (ML) techniques are widely used to interpret the collected IoT data ...for improving the company's operational excellence and predictive maintenance. In general, ML applications require high computational resource allocation and expertise. Manufacturing companies usually transfer their IIoT data to an ML-enabled third party or a cloud system. Although the transmission process uses encryption, ML applications still need decrypted data to perform ML tasks efficiently. Therefore, the third parties may have unacceptable access rights during the data processing to the content of IIoT data that contains a portrait of the production process. IIoT data may include hidden sensitive features, creating an information leakage for the companies. All these concerns prevent companies from sharing their IIoT data with third parties. This paper proposes a novel method based on the hybrid usage of Generative Adversarial Networks (GAN) and Differential Privacy (DP) to preserve sensitive data in IIoT operations. We aim to sustain IIoT data privacy with minimal accuracy loss without adding high additional computational costs to the overall data processing scheme. We demonstrate the efficiency of our approach with publicly available data sets and a realistic IIoT data set collected from a confectionery production process.
Computational capabilities of the largest high performance computing systems have increased by more than 100 folds in the last 10 years and keep increasing substantially every year. This increase is ...made possible mostly by multi-core technology besides the increase in clock speed of CPUs. Nowadays, there are systems with more than 100 thousand cores installed and available for processing simultaneously. Computational simulation tools are always in need of more than available computational sources. This is the case for especially complex, large scale flow problems. For these large scale problems, the soft error tolerance of the simulation codes should also be encountered where it is not an issue in relatively small scale problems due to the low occurrence probabilities. In this study, we analyzed the reaction of an incompressible flow solver to randomly generated soft errors at several levels of computation. Soft errors are induced into the final global assembly matrix of the solver by manipulating predetermined bit-flip operations. Behaviour of the computational fluid dynamics (CFD) solver is observed after iterative matrix solver, flow convergence and CFD iterations. Results show that the iterative solvers of CFD matrices are highly sensitive to customized soft errors while the final solutions seem more intact to bit-flip operations. But, the solutions might still differ from the real physical results depending on the bit-flip location and iteration number. So, the next generation computing platforms and codes should be designed to be able to detect bit-flip operations and be designed bit-flip resistant.
Graphic abstract
Full text
Available for:
CEKLJ, DOBA, EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, IZUM, KILJ, KISLJ, MFDPS, NLZOH, NUK, OILJ, PILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UILJ, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
Power flow calculations are crucial for the study of power systems, as they can be used to calculate bus voltage magnitudes and phase angles, as well as active and reactive power flows on lines. In ...this paper, a new approach, the Recycling Newton–Krylov (ReNK) algorithm, is proposed to solve the linear systems of equations in Newton–Raphson iterations. The proposed method uses the Generalized Conjugate Residuals with inner orthogonalization and deflated restarting (GCRO-DR) method within the Newton–Raphson algorithm and reuses the Krylov subspace information generated in previous Newton runs. We evaluate the performance of the proposed method over the traditional direct solver (LU) and iterative solvers (Generalized Minimal Residual Method (GMRES), the Biconjugate Gradient Stabilized Method (Bi-CGSTAB) and Quasi-Minimal Residual Method (QMR)) as the inner linear solver of the Newton–Raphson method. We use different test systems with a number of busses ranging from 300 to 70000 and compare the number of iterations of the inner linear solver (for iterative solvers) and the CPU times (for both direct and iterative solvers). We also test the performance of the ReNK algorithm for contingency analysis and for different load conditions to simulate optimization problems and observe possible performance gains.
•A novel recycling Krylov subspace approach for the power flow algorithm•Investigation of the angles between the Krylov subspaces generated from the Jacobians•ILU preconditioner is used once with AMD reordering for iterative solutions•Computational improvement due to similar subspaces for contingencies/varying loads•Extensive set of experiments for several different iterative and direct solvers.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
Beyin bilgisayar arayüzleri (BBA), beyin elektriksel aktivitelerini kontrol komutlarına çevirerek bilgisayar veya nöroprostetik kol gibi yardımcı teknolojilerin kullanılmasını sağlayan sistemlerdir. ...Bu çalışmada filtre tabanlı öznitelik seçim yöntemlerinin farklı sınıflandırma algoritmaları ile birlikte kullanılmasının BBA sistemlerine getirebileceği kazanımlar araştırılmıştır. Bu çerçevede nöroprostetik bir cihazın kontrolü için tasarlanan BBA sisteminden elde edilmiş EEG kayıtları analiz edilmiştir. EEG kayıtlarının analizi için delta (1.0-4 Hz), teta (4-8 Hz), alfa (8-12 Hz), beta (12-25 Hz), yüksek-beta (25-30Hz) ve gama (30-50 Hz) frekans bantlarından ve delta (1-4 Hz), teta (4-8 Hz), alfa1 (8-10 Hz), alfa2 (10-12 Hz), beta1 (12-15 Hz), beta2 (15-18 Hz), beta3 (18-25 Hz), gama1 (30-35 Hz), gama2 (35-40 Hz), gama3 (40-50 Hz) alt frekans bantlarından bant gücü öznitelikleri çıkarılmıştır. Elde edilen iki veri seti öznitelik seçimi uygulamadan ve öznitelik seçimi uygulayarak sınıflandırılmıştır. Çalışmada toplam 10 adet filtre tabanlı öznitelik seçimi yöntemi ile birlikte, doğrusal ayırt eden analizi, rassal ormanlar, karar ağaçları ve destek vektör makinaları sınıflandırma algoritmaları kullanılmıştır. Çalışma sonucunda EEG kayıtlarının sınıflandırılması için öznitelik seçme algoritmalarının uygulanmasının daha yüksek başarımlı sonuçlar verdiği ve bu çalışmada ele alınan öznitelik seçme yöntemlerinden, özdeğer merkeziyetine göre öznitelik seçimi (Ecfs) ve sonsuz öznitelik seçimi (Inffs) yöntemlerinin filtre tabanlı yaklaşımlar arasında en iyi sonuçları verdiği gözlenmiştir.
Full text
Available for:
IZUM, KILJ, NUK, PILJ, PNG, SAZU, UL, UM, UPUK
Quasi-static analysis of power systems can be performed by means of timeseries-based and probability density function-based models. In this paper, the effect of different load and generation ...modelling approaches on the quasi-static analysis of distribution networks is investigated. Different simplified load and distributed renewable energy sources generation timeseries-based models are considered as well as probabilistic analysis. Moreover, a more sophisticated approach based on cluster analysis is introduced to identify harmonized sets of representative load and generation patterns. To determine the optimum number of clusters, a three-step methodology is proposed. The examined cases include the quasi-static analysis of distribution networks for different operational conditions to identify the simplified modelling approaches that can efficiently predict the network voltages and losses. Finally, the computational efficiency by using the simplified models is evaluated in temperature-dependent power flow analysis of distribution networks.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
7.
Generic dynamic load modelling using cluster analysis Barzegkar-Ntovom, Georgios A.; Ceylan, Oguzhan; Papadopoulos, Theofilos A. ...
2018 53rd International Universities Power Engineering Conference (UPEC),
2018-Sept.
Conference Proceeding
In this paper, a new generic load modelling procedure is proposed, based on the application of cluster analysis on load model parameters identified from measured dynamic responses. The performance of ...the proposed approach is assessed using measurements obtained from a low-voltage laboratory scale test configuration. In order to develop robust generalized load models applicable to a wide range of operating conditions, different load compositions, operating conditions and voltage disturbances are considered in the analysis. The findings of this paper verify the validity of the proposed generic modelling procedure and indicate robust results using the proposed methodology.
This paper develops a model to show the effects of soft errors on power flow calculations. We create artificial bit-flips and apply them to solve linear equations obtained in Fast Decoupled Load Flow ...(FDLF) method. With this, we aim to represent the sensitivity of FDLF iteration against soft errors with a statistical view by using different numerical fault injection scenarios. The study focuses on the soft-error sensitivity of the FDLF method with exact (LU) linear solvers. We performed approximately 60000 experiments on IEEE 14, 30, 57, 118 test systems to examine the soft-error resiliency of FDLF method. Statistical study shows that even for small systems, soft-errors may have large impacts on the convergence of the FDLF. Hence, an improvement should be provided against soft errors in power flow studies to obtain more efficient and reliable simulations in near future exascale computational environments.
There are emerging trends to use the Industrial Internet of Things (IIoT) in manufacturing and related industries. Machine Learning (ML) techniques are widely used to interpret the collected IoT data ...for improving the company's operational excellence and predictive maintenance. In general, ML applications require high computational resource allocation and expertise. Manufacturing companies usually transfer their IIoT data to an ML-enabled third party or a cloud system. ML applications need decrypted data to perform ML tasks efficiently. Therefore, the third parties may have unacceptable access rights during the data processing to the content of IIoT data that contains a portrait of the production process. IIoT data may include hidden sensitive features, creating information leakage for the companies. All these concerns prevent companies from sharing their IIoT data with third parties. This paper proposes a novel method based on the hybrid usage of Generative Adversarial Networks (GAN) and Differential Privacy (DP) to preserve sensitive data in IIoT operations. We aim to sustain IIoT data privacy with minimal accuracy loss without adding high additional computational costs to the overall data processing scheme. We demonstrate the efficiency of our approach with publicly available data sets and a realistic IIoT data set collected from a confectionery production process. We employed well-known privacy six assessment metrics from the literature and measured the efficiency of the proposed technique. We showed, with the help of experiments, that the proposed method preserves the privacy of the data while keeping the Linear Regression (LR) algorithms stable in terms of the R-Squared accuracy metric. The model also ensures privacy protection for hidden sensitive data. In this way, the method prevents the production of hidden sensitive data from the sub-feature sets.
This paper proposes a methodology to characterize active and reactive power load profiles. Specifically, the approach makes use of fast Fourier Transform for conversion into frequency domain, ...principle component analysis to reduce the dimension and K-means++ to determine the representative load profiles. The data set consists of five-year measurements taken from the Democritus University of Thrace Campus. Test days were also classified as working and non-working. From the results it is observed that the proposed methodology determines representative load profiles effectively both regarding active and reactive power.