Data Mining Practical Machine Learning Tools and Techniques

https://bura.brunel.ac.uk/bitstream/2438/13152/4/Fulltext.pdf

Critical analysis of Big Data challenges and analytical methods

Giant and isotropic magnetoresistance as huge as −53% was observed in magnetic manganese oxide La0.72Ca0.25MnOz films with an intrinsic antiferromagnetic spin structure. We ascribe this magnetoresistance to spin‐dependent electron scattering due to spin canting of the manganese oxide.

31p-S-4 Giant Magnetoresistance in Magnetic Manganese Oxide with Intrinsic Antiferromagnetic Spin Structure

In the era of the Internet of Things (IoT), an enormous amount of sensing devices collect and/or generate various sensory data over time for a wide range of fields and applications. Based on the nature of the application, these devices will result in big or fast/real-time data streams. Applying analytics over such data streams to discover new information, predict future insights, and make control decisions is a crucial process that makes IoT a worthy paradigm for businesses and a quality-of-life improving technology. In this paper, we provide a thorough overview on using a class of advanced machine learning techniques, namely deep learning (DL), to facilitate the analytics and learning in the IoT domain. We start by articulating IoT data characteristics and identifying two major treatments for IoT data from a machine learning perspective, namely IoT big data analytics and IoT streaming data analytics. We also discuss why DL is a promising approach to achieve the desired analytics in these types of data and applications. The potential of using emerging DL techniques for IoT data analytics are then discussed, and its promises and challenges are introduced. We present a comprehensive background on different DL architectures and algorithms. We also analyze and summarize major reported research attempts that leveraged DL in the IoT domain. The smart IoT devices that have incorporated DL in their intelligence background are also discussed. DL implementation approaches on the fog and cloud centers in support of IoT applications are also surveyed. Finally, we shed light on some challenges and potential directions for future research. At the end of each section, we highlight the lessons learned based on our experiments and review of the recent literature.

Deep Learning for IoT Big Data and Streaming Analytics: A Survey

Healthcare data are a valuable source of healthcare intelligence. Sharing of healthcare data is one essential step to make healthcare system smarter and improve the quality of healthcare service. Healthcare data, one personal asset of patient, should be owned and controlled by patient, instead of being scattered in different healthcare systems, which prevents data sharing and puts patient privacy at risks. Blockchain is demonstrated in the financial field that trusted, auditable computing is possible using a decentralized network of peers accompanied by a public ledger. In this paper, we proposed an App (called Healthcare Data Gateway (HGD)) architecture based on blockchain to enable patient to own, control and share their own data easily and securely without violating privacy, which provides a new potential way to improve the intelligence of healthcare systems while keeping patient data private. Our proposed purpose-centric access model ensures patient own and control their healthcare data; simple unified Indicator-Centric Schema (ICS) makes it possible to organize all kinds of personal healthcare data practically and easily. We also point out that MPC (Secure Multi-Party Computing) is one promising solution to enable untrusted third-party to conduct computation over patient data without violating privacy.

Healthcare Data Gateways: Found Healthcare Intelligence on Blockchain with Novel Privacy Risk Control

Angular resolution is crucial for a scanning radar system since it is limited by real aperture. Aiming at this problem, a scheme of angular superresolution for scanning radar with MUSIC based on virtual array mapping is proposed. First, according to the similarity between the power antenna adopted in scanning radar and that in uniform linear array (ULA) processing, the azimuth echo is mapped to the common beamformer (CBF) output of a corresponding virtual ULA. Then based on this mapping relationship, the data covariance matrix of the virtual ULA is estimated by using the least squares (LS) criterion and the targets distribution is recovered via multiple signal classification (MUSIC) algorithm. Simulation validates that the method can improve the scanning radar angular resolution.

Angular superresolution for scanning radar with MUSIC based on virtual array mapping

This paper presents an adjustable angular superresolution method to realize high azimuthal resolution of forward-looking area in scanning radar imaging. Firstly, the received signal in azimuth dimension is established as the convolution model of target scattering coefficient and antenna pattern. Then based on the Poisson statistic assumption, the Generalized Gaussian distribution as prior constraint is used in the maximum a posterior (MAP) method due to the adjustability of dispersion parameter. At last, how to choose suitable dispersion parameter is discussed for better superresolution performance of different scenes. The simulations and experimental results are given to verify the performance of proposed method.

Bayesian Superresolution Method of Forward-Looking Imaging with Generalized Gaussian Constraint

LTE-based humanoid robotics system

The high azimuth angular resolution problem of real-beam scanning radar is significant to targets detection and location. ML iterative adaptive approach(ML-IAA) has been used in array signal processing to realize high angular resolution. In order to improve the azimuth angular resolution of real-beam scanning radar, we introduce this algorithm to real-beam radar system, called real-beam ML iterative superresolution approach(RML-ISA). This method established the likelihood function by utilizing the statistical property of real beam data. Applying this method to the real-beam radar system only needs few scanning echo to obtain effective results. Simulations illustrate the performance of our algorithm.

ML iterative superresolution approach for real-beam radar

Synthetic aperture radar automatic target recogni­tion (SAR ATR) can be benefited from multi-view observation. It is important to consider the tradeoff between the recognition performance of the multi-view ATR system and time efficiency during reconnaissance. Here, we propose a new multi-view ATR algorithm to realize the optimal target recognition meanwhile guaranteeing the SAR platform with a short flight path. The basic scheme of the multi-view classifier based on ensemble learning is proposed, and the convolutional neural networks with two inputs are trained as the base classifiers. A directed graph forms to represent the tradeoff between the recognition performance and flight distance. Thus the optimal viewpoint selection for multi-view SAR ATR is solved by the single source shortest path search algorithm in directed graph. The multi-view SAR images are collected on the solved optimal viewpoints, and the final recognition result is obtained by the base classifiers ensemble. Some experiments based on MSTAR dataset have shown the superiority of the proposed method both on optimal viewpoint selection and recognition performance.

Yin Zhang

Papers

Angular superresolution for scanning radar with MUSIC based on virtual array mapping

Bayesian Superresolution Method of Forward-Looking Imaging with Generalized Gaussian Constraint

LTE-based humanoid robotics system

ML iterative superresolution approach for real-beam radar

Multi-view SAR ATR based on networks ensemble and graph search