Effective Statistical Methods for Big Data Analytics

With advances in technologies in the past decade, the amount of data generated and recorded has grown enormously in virtually all fields of industry and science. This extraordinary amount of data provides unprecedented opportunities for data-driven decision-making and knowledge discovery. However, the task of analyzing such large-scale dataset poses significant challenges and calls for innovative statistical methods specifically designed for faster speed and higher efficiency. In this chapter, we review currently available methods for big data, with a focus on the subsampling methods using statistical leveraging and divide and conquer methods.

Design and Analysis of Computer Experiments

Computer experiments refer to the study of complex systems using mathematical models and computer simulations. The use of computer experiments becomes popular for studying complex systems in science and engineering. The design and analysis of computer experiments have received broad attention in the past decades. In this chapter, we present several widely used statistical approaches for design and analysis of computer experiments, including space-filling designs and Gaussian process modeling. A special emphasis is given to recently developed design and modeling techniques for computer experiments with quantitative and qualitative factors.

Supervised Learning in Absence of Accurate Class Labels

Traditionally supervised learning algorithms are built using labeled training data. Accurate labels are essential to guide the classifier towards an optimal separation between the classes. However, there are several real world scenarios where the class labels at an instance level may be unavailable or imprecise or difficult to obtain, or in situations where the problem is naturally posed as one of classifying instance groups. To tackle these challenges, we draw your attention towards Multi Instance Learning (MIL) algorithms where labels are available at a bag level rather than at an instance level. In this chapter, we motivate the need for MIL algorithms and describe an ensemble based method, wherein the members of the ensemble are lazy learning classifiers using the Citation Nearest Neighbour method. Diversity among the ensemble methods is achieved by optimizing their parameters using a multi-objective optimization method, with the objective being to maximize positive class accuracy and minimize false positive rate. We demonstrate results of the methodology on the standard Musk 1 dataset.

Statistical Analysis of Functional Magnetic Resonance Imaging Data

The analysis of functional magnetic resonance imaging (fMRI) data poses many statistical challenges. The data are massive, noisy, and have a complicated spatial and temporal correlation structure. This chapter introduces the basics of fMRI data collection and surveys common approaches for data analysis.

Cyber-Physical Systems

Pervasive computing has progressed significantly with a growth of embedded systems as a result of recent advances in digital electronics, wireless networking, sensors and RFID technology. These embedded systems are capable of producing enormous amount of data that cannot be handled by human brains. At the same time, there is a growing need for integrating these embedded devices into physical environment in order to achieve a far better capability, scalability, resiliency, safety, security and usability in important sectors such as healthcare, manufacturing, transportation, energy, agriculture, architecture and many more. The confluence of all these recent trends is the vision of distributed cyber-physical systems that will far exceed the performance of traditional embedded systems. Cyber-physical systems are emerging technology that require significant research in design and implementation with a few important challenges to overcome. The goal of this chapter is to present an overview of basic design and architecture of a cyber-physical system along with some specific applications and a brief description of the design process for developers. This chapter also presents a brief discussion of security and privacy issues, the most important challenge of cyber-physical systems.

Support Vector Machines and Applications

Support Vector Machines is one of the powerful Machine learning algorithms used for numerous applications. Support Vector Machines generate decision boundary between two classes which is characterized by special subset of the training data called as Support Vectors. The advantage of support vector machine over perceptron is that it generates a unique decision boundary with maximum margin. Kernalized version makes it very faster to learn as the data transformation is implicit. Object recognition using multiclass SVM is discussed in the chapter. The experiment uses histogram of visual words and multiclass SVM for image classification.

Machine Learning Methods as a Test Bed for EEG Analysis in BCI Paradigms

Machine learning techniques, is a crucial tool to build analytical models in EEG data analysis. These models are an excellent choice for analyzing the high variability in EEG signals. The advancement in EEG-based Brain-Computer Interfaces (BCI) demands advanced processing tools and algorithms for exploration of EEG signals. In the context of the EEG-based BCI for speech communication, few classification and clustering techniques is presented in this book chapter. A broad perspective of the techniques and implementation of the weighted k-Nearest Neighbor (k-NN), Support vector machine (SVM), Decision Tree (DT) and Random Forest (RF) is explained and their usage in EEG signal analysis is mentioned. We suggest that these machine learning techniques provides not only potentially valuable control mechanism for BCI but also a deeper understanding of neuropathological mechanisms underlying the brain in ways that are not possible by conventional linear analysis.

Modeling Associations

In this work, the authors develop signaling models based on ideas from homology and discuss how to design a model of signal space which decomposes the incoming signals into classes of progressively higher levels of associative meaning. The tools needed are illustrated with a simple approach using standard linear algebra processing but this is just a simple point of departure into a more complex and potentially more useful signal processing toolbox involving computational homology. These ideas then lead to models of grammar for signals in terms of cascaded barcode signal representations.

Measuring Complexity of Chaotic Systems With Cybernetics Applications

Measuring complexity of systems is very important in Cybernetics. An aging human heart has a lower complexity than that of a younger one indicating a higher risk of cardiovascular diseases, pseudo-random sequences used in secure information storage and transmission systems are designed to have high complexity (to resist malicious attacks), brain networks in schizophrenia patients have lower complexity than corresponding networks in a healthy human brain. Such systems are typically modeled as deterministic nonlinear (chaotic) system which is further corrupted with stochastic noise (Gaussian or uniform distribution). After briefly reviewing various complexity measures, this chapter explores characterizing the complexity of deterministic nonlinear chaotic systems (tent, logistic and Hénon maps, Lorenz and Rössler flows) using specific measures such as Lempel-Ziv complexity, Approximate Entropy and Effort-To-Compress. Practical applications to neuron firing model, intra-cranial pressure monitoring, and cardiac aging detection are indicated.

Patient Data De-Identification

With the rapid increment in the clinical text, de-identification of patient Protected Health Information (PHI) has drawn significant attention in recent past. This aims for automatic identification and removal of the patient Protected Health Information from medical records. This paper proposes a supervised machine learning technique for solving the problem of patient data de- identification. In the current paper, we provide an insight into the de-identification task, its major challenges, techniques to address challenges, detailed analysis of the results and direction of future improvement. We extract several features by studying the properties of the datasets and the domain. We build our model based on the 2014 i2b2 (Informatics for Integrating Biology to the Bedside) de-identification challenge. Experiments show that the proposed system is highly accurate in de-identification of the medical records. The system achieves the final recall, precision and F-score of 95.69%, 99.31%, and 97.46%, respectively.