Deep Learning and Edge Computing Solutions for High Performance Computing

Deep learning is a promising way to get relevant information from IoT service sensor data embedded in complex situations. Due to its multifaceted structure, deep learning is better suited to the nature of computer drag. So, in the course of this article, we start by introducing deep IoT metrics into the computer environment. Because there is a limited amount of available bandwidth, we are designing a separate load-filling strategy to optimize the performance of deep IoT learning systems using computers (World Health Organization Epilepsy, http://​www.​who.​int/​mediacentre/​factsheets/​fs999/​en/​)

With the abundance of novel diagnostic techniques and digital storage methodologies, a huge amount of big data is available in the healthcare sector across the globe. This information can be used judiciously for drawing proper conclusions with respect to diagnosis, treatment, and prevention of a disease state. The major hold back in the analysis of the data is the interest of privacy preservation. Hence public availability of data is not supported by the ethical medical community and legal authorities. The data stored in the databases contribute to big data not only in their volume but also with respect to variety and velocity with which they are generated and stored. The origin of information from healthcare databases can serve subsequent applications like drug development, disease diagnosis, and so on. Artificial Intelligence, on the other hand, provides analytic techniques for drawing patterns and anomalies present in the databases. With the support of data-mining and machine learning techniques, AI can be used effectively for grabbing the hidden information stored in the healthcare databases even while the integrity and privacy of data are maintained. These healthcare databases include data in the form of images, text, and statistics that are looked upon by techniques of AI in data analysis. Data mining focuses on identifying unknown patterns and building models for prediction that can assist the medical society for disease discovery and to suggest probable treatment schedule. Machine learning techniques will look into the possibility of reinforcement learning for assisting the pharma industry for drug development and the medical team for generating treatment plans. We expect the healthcare industry to reap the analytical capabilities of AI techniques for delivering prompt and timely diagnosis and treatment.

Recently, dengue has become a worldwide phenomenon and is one of the major public health concerns. There are lots of areas that are affected by dengue in India. West Bengal is one of the states in India which ranks at a higher level among the dengue-affected people in India. Among the various districts of West Bengal, Kolkata occupies the first rank in dengue-affected areas. In this paper, an attempt has been made to find out the spatiotemporal pattern of dengue disease in Kolkata using the geographical information system (GIS) as it is one of the latest approaches in healthcare informatics. The objective of this study is to find out the spatiotemporal pattern and hotspot identification of dengue-affected areas in Kolkata. In this study, analysis has been done using hotspot identification of the dengue-affected areas. This study reveals that spatiotemporal diffusion pattern of the dengue disease in Kolkata city. By the borough level analysis, it has been found that the outbreaks of dengue diseases are mostly concentrated in the central and southern part of Kolkata. This study represents useful information about the dengue-affected areas of Kolkata city. It is also helpful to the health departments of West Bengal and policymakers to make planning strategies to prevent dengue. The methodology used in this dengue study can also be applied to other studies like malaria, influenza, and so on.

The electronic computer is one of the most important developments of the twentieth century. Almost all kinds of people are getting the benefit of computing for one or another purpose in life. Since the inception of the computer, there is a paradigm shift of computing from mainframes to microcomputers, microcomputers to a personal computer (pc), pc to centralized (web), and then to the cloud. This chapter outlines the term computing and its purpose in everyday life. It even explained what comprises a computing environment and mentioned the various types of computing environments. Moreover, the market size of edge computing by value is presented, and various players in the market are mentioned. The demands of micro edge data centers have been increasing in the market mainly to scale up the growth of edge computing platform-based applications as opposed to conventional data center environments. Furthermore, the chapter explains the concept of edge computing and how it differs from cloud computing. In addition, typical layered architecture of the edge computing ecosystem is presented, and the various types of components present in each layer are outlined. The characteristics of edge computing and the various types of edge computing devices are also stated. Besides, the main benefits of edge computing are also briefly explained. Finally, two use cases where edge computing brings new values to the society namely IoT-based automated assisted living care and contactless healthcare services are also discussed.

Proliferation of social networking sites, IoT, and a huge number of electronic transactions have given rise to a data deluge. This huge amount of data combined with cloud storage and proliferation of graphics processing units (GPUs) have ushered in a new era of machine learning (ML) and deep learning (DL). These techniques have been very useful in analyzing the data quickly in a wide range of applications such as self-driving cars, virtual reality, robotics, healthcare, and so on. However, resource-constrained end devices may not be suitable for computationally intensive operations that deep learning demands. Processing and analysis of the data can be done on the cloud but will involve high bandwidth usage, latency in obtaining the results, and also privacy concerns that are not acceptable for these applications. One alternative is to use edge computing that keeps the data in situ and brings the applications in close proximity to the data (i.e., at the network edge) so as to reduce the communication cost, lower the bandwidth and latency, and also, adds a layer of security to the data. A paradigm that complements cloud computing and benefits each other is Edge Computing.

Ingenious situation monitoring and defect pronouncement by examining sensor data that can guarantee the invulnerability of machinery. Accustomed defect diagnosis and assessment method formally enforce certain conducts to bring down the noise and derives some time province or regularity province aspects from antenna data allied with unrefined moment series. Then some strategies are implemented to make analyzation. However, these accustomed defect verdict approaches have an inadequacy with the feature selection and also discard the consideration of temporal logic of time series data. The aforementioned paper contemplates a defect diagnosis model based on Deep Stack Neural Networks (DSNN). This is an exemplary model that can precisely diagnose the raw sensor data allied with time series neglecting the tedious selection process and indicator processing. It also avails the benefit of facilitating the temporal rationality of the data. Primary essence of the process is to diminish the cost function value. In order to acquire the minimal value the raw time series training data are levelheaded by the sensors. The next level is to scrutinize the classification accuracy of the Deep Stack Neural Networks with times series data achieved formerly. Conclusively defect diagnosis with former data allied with time series is implemented including the temporal coherence. Consequently, a complete neglecting of defects and prominent aftereffect of achieving a fault bearing classification accuracy. Thus the set forth paper assures an effective identity of defect bearing system.

Position control applications like robotic arms, CNC machines, and so on, can perform better with the use of 2 phase AC servo motor. The exceptionally nonlinear attributes of this motor have presented issues for its high accuracy in position control. In addition, a periodic motion makes a control system lose its function due to accumulating errors with an increasing number of periods. Although the servo tracking efficiency is good, but it declines periodic time varying signals. So for the better operations of the AC servo motor in the position control, a Fuzzy Adaptive Intelligent Controller (FAIC) is attempted to enhance the periodic reference tracking. The experimental study is conducted with AC servo motor system as shown in Fig. 1 for the identification of the process dynamics. Simulation results obtained to authenticate the proposed controller delivered good tracking performance than the Conventional fuzzy controller and conventional PD controller. For periodic reference tracking, the proposed controller exhibits a minimum Absolute Tracking Error (ATE) profile is the least with FAIC when compared to other controllers. The authors have also tested the robustness of the control strategy.

Healthcare sectors of different types and specialties are increasing day by day and kept as a data pool called Electronic health records. This pool of data that healthcare owned is also enormously increasing and becomes unused without the intervention of a proper trained model for further medical analysis. Artificial Intelligence, machine learning, and deep learning are some of the known technologies that can help the end user to analyze and provide recommendations. Through the advent of this innovative technology, hierarchical learning, or deep structured learning is going to be done on the existing EHRs using layered algorithmic architecture for a complete data analysis within a short duration of time. With successful experimental results and enormous applications areas, deep learning has the potential to change the challenges pertaining to healthcare.

Deep learning is a much focused domain of artificial neural networks. Deep learning algorithms try to learn massive amounts of unlabelled data and make a better analysis. With deep learning, all layers learn the input data and transform it into a more abstract and composite format. The word “deep” means higher numbers of hidden layers in which the data from one layer to another is transformed to generate the most accurate outcome. Deep learning architecture has been applied to different fields like medical image analysis, machine translation, bioinformatics, speech recognition, social network filtering, computer vision, audio recognition drug design, natural language processing, and so on. This chapter discusses important deep learning applications across different disciplines, their contribution to the real world, and a study of the architectures and methods used by each application. This chapter also introduces the differences between machine learning and deep learning. Finally, this chapter concludes with future aspects and conclusions.

Deep Learning (DL), a division of Machine Learning (ML) is a highly focused field of data science. DL is the most active approach for ML. DL algorithms excerpt the complex high-level data features via a hierarchical learning process. These complex abstractions at a given hierarchical level are learned based on the abstractions framed in the previous level of the hierarchy. The capability of DL to analyze and learn huge quantities of unsupervised data makes it a powerful tool for Big Data Analytics where data are mostly unorganized and unlabeled. With the significant advancements and tremendous performance of DL, it is broadly used across numerous domains such as business, health, government, and so on. This chapter focuses on the overview and applications of DL from a real-world perspective, which covers a variety of areas such as Speech Recognition, Text Classification, Document Summarization, Fraud Detection, Visual Recognition, Personalization’s, and so on.

Agriculture is one of the world’s most influential sectors. Agricultural productivity is beneficial to the economy, privacy, nutrition, and health of a country. In the modern era, agronomists have taken various technologies like IoT and Blockchain to get better farming crops. Blockchain technology used to address several problems from various fields like economy, health, and energy. Blockchain utilized in supply chain management structures in the agriculture sectors that provide accountability, protection, neutrality, and efficiency for all processes in a supply chain. It helps to solve many of the internet’s stability and security complications. In this chapter, we deliberate the use case of blockchain and its functionality process in agriculture. Further explores blockchain technology, application, challenges, and opportunities toward the field of agriculture.

Deep learning (DL) techniques have recently emerged as the most significant techniques for processing big multimedia data. DL networks autonomously extract advanced and inherent features from the big data sets using systematic learning methods. The real-world problem-solving using DL techniques demands large parallel computing infrastructure facilities for achieving high efficiency. Recent developments in deep learning techniques have demonstrated that it could outperform humans in some tasks such as classifying and tracking multimedia data. The deep learning networks can have about 150 hidden layers. The increase in the output performance of deep learning networks is directly proportional to input sample data size. This paper reviewed the literature on applications on deep learning from diverse application domains. Authors have also carried out a comparative study of various DL methods used and highlighted their results.

Health informatics or healthcare informatics is a field of science that is evolving with the expansion of electronic health records (EHRs) and health analytics systems. At present, several applications are developed where the doctor is aided by the machine in detecting abnormalities to provide better healthcare. Here the focus is on EHRs and health analytics systems in detecting diabetic retinopathy and coronary heart disease identification. The ultimate aim is to identify the diverse forms of lesions during the early stages discovery of diabetic retinopathy. Search early detection could help to prevent permanent vision loss among diabetic patients. Image processing techniques that detect diabetic retinopathy lesions, aid in examining the image of the damaged part of the retina. The machine learning algorithm used employees’ specific color channels image features to separate physiological features from exudates digitally. The five-stage classification of the severity of the disease comprises of three stages of low risk and two stages of diabetic retinopathy. By implementing an automated system for identification of this disease we have a chance to accurately detect an affected patient easily. The second objective which is an intelligent system to detect coronary heart disease involves monitoring the heartbeat signals defined variations and monitoring patient’s health status automatically through sensor-dependent networks that employ internet of things technology. The primary aim here is that the patient is constantly monitored. This system relies on wireless sensors that are used the monitor cardiac patients using piezoelectric sensors that are utilized for measuring the arteries’ thickness and extract the waveforms without any intervention from humans. These collected waveforms can then be classified as normal or abnormal. This information is communicated to a mobile app immediately. The mobile plays the role of a display device and is also capable of uploading data to a cloud platform for detailed analysis. A cardiologist can then access the data and results from the cloud database.

Malaria is a migratory and easily transmissible disease transmitted by virus carrier mosquitoes. Visual quantification and classification of parasitemia in emaciated plasma films is a very tiresome, biased, and time-consuming task. This chapter delivers an unambiguous understanding of the quantification and classification of erythrocytes in tainted wafer-thin blood films diseased with plasmodium parasites. The approach mainly includes microscopic imaging of tainted blood glides, amputation of noise and illumination adjustment, erythrocyte segmentation, morphological operations. By means of the segmented illustrations, parasite density estimation and classification of the stage of infection are done. Two different classification techniques ANN and DCNN have been designed and tested to perform the grouping of diseased erythrocytes into their corresponding stages of growth. The traditional neural network ANN approach gave an accuracy of 93% and this was again overwhelmed by using customized Deep Convolutional Network by achieving an accuracy of 95%.

High-performance computing (HPC) is achieved from parallel processing and nowadays, it can also be achieved from distributed processing techniques for solving complex compute-intensive and data-intensive applications. HPC tends to solve complex problems in less time and efficiently with parallel processing techniques. Computer modeling and research activities can be carried out with the help of HPC. It is used for solving contemporary issues. HPC makes efficient utilization of the available resources to provide unremitting performance. HPC can be combined with deep learning techniques to get superior performance. This chapter discusses distributed and parallel deep learning and their applications in the real world.