China
Africa
Explore chapters and articles related to this topic
Introduction
Published in Bhavani Thuraisngham, Murat Kantarcioglu, Latifur Khan, Secure Data Science, 2022
Bhavani Thuraisngham, Murat Kantarcioglu, Latifur Khan
Part III, consisting of five chapters, will discuss security and privacy-enhanced data science. Chapter 10 discusses attacks to data science systems and describes our approach to adversarial machine learning in general and support vector machine learning in particular. Chapter 11 describes adversarial relevance vector machine learning. Privacy-enhanced data science with decision trees as an example is discussed in Chapter 12. Finally, our work on a privacy-aware policy-based data management framework for the quantified self-applications is discussed in Chapter 13. Finally, in Chapter 14, we discuss how our work can be applied to a specific application and that is the COVID-19 pandemic.
Fuzzy Relevance Vector Machines with Application to Surface Electromyographic Signal Classification
Published in Biju Issac, Nauman Israr, Case Studies in Intelligent Computing, 2014
Hong-Bo Xie, Hu Huang, Socrates Dokos
To overcome these problems of SVM eciently, Tipping [21] developed a new kernel-based machine learning technique, termed relevance vector machine (RVM). e RVM shares many of the characteristics of the SVM while avoiding its principal limitations. It uses the sparse Bayesian learning framework, in which an a priori parameter structure is placed based on automatic relevance determination theory for removing irrelevant data points [22]. Hence, it produces sparse models as well as a comparable generalization performance to that of the SVM. Most importantly, RVM classication requires dramatically fewer relevance vectors (RVs) compared with the number of SVs for SVM classication. is can signicantly reduce the computational cost, making the RVM more suitable for real-time applications [23-26]. Many EMG-based controls, including prosthetic hand and exoskeletal control, as well as wheelchair and robotic control, need to be performed in real time [2,3,5,6,9,13]. e RVM is thus a potentially promising tool to classify EMG patterns. Similar to the SVM, the original RVM is a binary classier. As for multiclass recognition, several coding schemes have been proposed using binary classiers [18,21]. However, indecisive regions often exist when a binary RVM classier ensemble is used to accommodate a multiclass problem (explained in Section 8.2). In order to solve for unclassiable regions in the RVM, we dene two membership functions in a direction perpendicular to the optimal hyperplane that separates the pair of classes. We evaluate the performance of the proposed fuzzy relevance vector machines (FRVMs) in classication of six hand motions using four-channel EMG signals. Fuzzy support vector machines (FSVMs) based on a leastsquares algorithm are compared in terms of classication accuracy, sparsity, and processing delay.
Displacement prediction of water-induced landslides using a recurrent deep learning model
Published in European Journal of Environmental and Civil Engineering, 2023
Qingxiang Meng, Huanling Wang, Mingjie He, Jinjian Gu, Jian Qi, Lanlan Yang
Compared with the physical models, data-driven models are intuitive and widely applied in displacement prediction. The evolution of a landslide is predicted with the use of monitoring data. Based on analytical tools, these methods can be divided into two categories. The first category is statistical methods using traditional theories such as regression models (Rose & Hungr, 2007), fractal models (Xu et al., 2016), nonlinear dynamics models (Qin et al., 2002), etc. In these methods, the evolution of landslides is assumed in the form of a certain mathematical model. The models often concentrate on the main feature of a displacement time-series and neglect other information. Another group of data-driven models is the artificial intelligence-based approaches. With the rapid development of computational intelligence, a variety of machine-learning models have been applied to landslide displacement prediction. Du et al. (2013) adopted a back-propagation neural network model for prediction of periodic displacement. Miao et al. (2018) proposed a support vector regression model. Liu et al. (2014) proposed a study that compared the different computational intelligence approaches of support vector machine (SVM), Gaussian process (GP), relevance vector machine (RVM) and simple artificial neural network (ANN). Recently, the extreme learning machine (ELM) has become a popular method for landslide displacement prediction (Li et al., 2018, Zhou et al., 2018).
Machine learning algorithms applied to intelligent tyre manufacturing
Published in International Journal of Computer Integrated Manufacturing, 2023
Simone Massulini Acosta, Rodrigo Marcel Araujo Oliveira, Ângelo Márcio Oliveira Sant’Anna
Onan and Toçoglu (2021) used a long short-term memory learning algorithm to identify sarcastic text documents. The results demonstrated that this deep learning algorithm could be used in natural language processing for the sarcasm identification task. Acosta et al. (2021) combined a relevance vector machine algorithm with a novel self-adaptive differential evolution metaheuristic for predictive modeling in a steelmaking process. This study described the performance of the proposed algorithm with the random forest, artificial neural network, and k-nearest neighbors algorithms. Farahani et al. (2021) used ten machine learning algorithms to analyze quality monitoring systems from in-mold sensors and machine data. They proposed an approach to predict the best weight, diameter, and thickness value from injection molding processes. Acosta and Sant’anna (2022) presented an approach for monitoring product failures based on machine learning algorithms integrated with the control chart method. The results showed that the relevance vector machine performs better than the support vector machine and artificial neural network.
A Novel Machine Learning Based Framework for Detection of Autism Spectrum Disorder (ASD)
Published in Applied Artificial Intelligence, 2022
Hamza Sharif, Rizwan Ahmed Khan
One of the promising studies done by Sabuncu et al. (Sabuncu et al. 2015) used the Multivariate Pattern Analysis (MVPA) algorithm and structural MRI (s-MRI) data to predict chain of neurodevelopmental disorders i.e. Alzheimer’s, Autism, and Schizophrenia. Sabuncu et al. analyzed structural neuroimaging data from six publicly available websites (https://www.nmr.mgh.harvard.edu/lab/mripredict), with 2800 subjects. The MVPA algorithm constituted with three classes of classifiers that includes a Support Vector Machine (SVM) (Vapnik 2013), Neighborhood Approximation Forest (NAF) (Konukoglu et al. 2012) and Relevance Vector Machine (RVM) (Tipping 2001). Sabuncu et al. attained detection accuracies of 70%, 86% and 59% for schizophrenia, Alzheimer and autism, respectively, using a 5-fold validation scheme (refer Section 4.1.3 for discussion on -fold cross validation methodology).