Towards Fine-Grained Online Adaptive Approximation Control for Dense SLAM on Embedded GPUs

Dense SLAM is an important application on an embedded environment. However, embedded platforms usually fail to provide enough computation resources for high-accuracy real-time dense SLAM, even with high-parallelism architecture such as GPUs. To tackle this problem, one solution is to design proper approximation techniques for dense SLAM on embedded GPUs. In this work, we propose two novel approximation techniques, critical data identification and redundant branch elimination. We also analyze the error characteristics of the other two techniques—loop skipping and thread approximation. Then, we propose SLaPP, an online adaptive approximation controller, which aims to control the error to be under an acceptable threshold. The evaluation shows SLaPP can achieve 2.0× performance speedup and 30% energy saving on average compared to the case without approximation.

Why Personal Dreams Matter: How professionals affectively engage with the promises surrounding data-driven healthcare in Europe

Recent buzzes around big data, data science and artificial intelligence portray a data-driven future for healthcare. As a response, Europe's key players have stimulated the use of big data technologies to make healthcare more efficient and effective. Critical Data Studies and Science and Technology Studies have developed many concepts to reflect on such overly positive narratives and conduct critical policy evaluations. In this study, we argue that there is also much to be learned from studying how professionals in the healthcare field affectively engage with this strong European narrative in concrete big data projects. We followed twelve hospital-based big data pilots in eight European countries and interviewed 145 professionals (including legal, governance and ethical experts, healthcare staff and data scientists) between 2018 and 2020. In this study, we introduce the metaphor of dreams to describe how professionals link the big data promises to their own frustrations, ideas, values and experiences with healthcare. Our research answers the question: how do professionals in concrete data-driven initiatives affectively engage with European Union's data hopes in their ‘dreams’ – and with what consequences? We describe the dreams of being seen, of timeliness, of connectedness and of being in control. Each of these dreams emphasizes certain aspects of the grand narrative of big data in Europe, makes particular assumptions and has different consequences. We argue that including attention to these dreams in our work could help shine an additional critical light on the big data developments and stimulate the development of responsible data-driven healthcare.

Digital Equity

This chapter discusses digital equity through the lens of the digital divide. While the digital divide is as old as information communication technology itself (ICT), the COVID-19 health crisis renewed a strident interest in exposing the significant gap that still exists after close to 30 years. The digital divide then is first contextualized within the coronavirus pandemic to illustrate how inequities came further to the forefront of people's agenda. It then moves to discuss the digital divide defining the complex term and offering critical data to illustrate the areas of the world most impacted by this unfortunate reality. Different organizations and groups have made significant moves to narrow the digital gap. These strategies are discussed next. None of these groups will be fully successful if, as will be argued, they are not concerned with digital equity. Finally, the chapter makes some critical observations on future challenges facing ICT vis-à-vis the digital divide.

MRI Image Based Classification Model for Lung Tumor Detection Using Convolutional Neural Networks

Lung tumor is a dangerous disease with the most noteworthy effects and causing more deaths around the world. Medical diagnosis of lung tumor growth can essentially lessen the death rate, on the grounds that powerful treatment alternatives firmly rely upon the particular phase of disease. Medical diagnosis considers to the use of innovation in science with the end goal of analyzing the interior structure of the organs of the human body. It is an approach to improve the nature of the patient's life through a progressively exact and fast detection, and with restricted symptoms, prompting a powerful generally treatment methodology. The main goal of the proposed work is to design a Lung Tumor Detection Model using Convolution Neural Networks (LTD-CNN) with machine learning technique that spread both miniaturized scale and full scale image surfaces experienced in Magnetic Resonance Imaging (MRI) and advanced microscopy modalities separately. Image pixels can give critical data on the abnormality of tissue and performs classification for accurate tumor detection. The advancement of Computer-Aided Diagnosing (CAD) helps the doctors and radiologists to analyze the lung disease precisely from CT images in its beginning phase. Different methods are accessible for the lung disease recognition, however numerous methodologies give not so much exactness but rather more fake positives. The proposed method is compared with the traditional models and the results exhibit that the proposed model detects the tumor effectively and more accurately.

PEDTARA: Priority-Based Energy Efficient, Delay and Temperature Aware Routing Algorithm Using Multi-Objective Genetic Chaotic Spider Monkey Optimization for Critical Data Transmission in WBANs

Software-Defined Wireless Body Area Network (WBAN)s have gained significance in emergency healthcare applications for remote patients. Prioritization of healthcare data traffic has a high influence on the congestion and delay in the WBAN routing process. Currently, the energy constraints, packet loss, retransmission delay and increased sensor heat are pivotal research challenges in WBAN. These challenges also degrade the network lifetime and create serious issues for critical health data transmission. In this context, a Priority-based Energy-efficient, Delay and Temperature Aware Routing Algorithm (PEDTARA) is presented in this paper using a hybrid optimization algorithm of Multi-objective Genetic Chaotic Spider Monkey Optimization (MGCSMO). This proposed optimized routing algorithm is designed by incorporating the benefits of chaotic and genetic operators to the position updating function of enhanced Spider Monkey Optimization. For the prioritized routing process, initially, the patient data transmission in the WBAN is categorized into normal, on-demand and emergency data transmissions. Each category is ensured with efficient routing using the three different strategies of the suggested PEDTARA. PEDTARA performs optimal shortest path routing for normal data, energy-efficient emergency routing for high priority critical data and faster but priority verified routing for on-demand data. Thus, the proposed PEDTARA ensures energy-efficient, congestion-controlled and delay and temperature aware routing at any given period of health monitoring. Experiments were performed over a high-performance simulation scenario and the evaluation results showed that the proposed PEDTARA performs efficient routing better than the traditional approaches in terms of energy, temperature, delay, congestion and network lifetime.

A Success Story of Critical Data Gathering During the Development Phase of Extreme ERD Well Drilling

This paper discusses a holistic combination of advanced formation evaluation techniques with pressure testing and reservoir navigation services to mitigate uncertainty related challenges in real time and successfully drill & place ERD laterals targeting Jurassic carbonate reservoirs. A meticulously planned approach to navigate the well trajectory by tracking the desired properties, informed decision-making while drilling and accurate data acquisition for aiding appropriate selection and placement in-flow control device (ICD) in lower completion design and future reservoir management contributed to the success of these complex wells in carbonate reservoirs. The first well in this study, involved drilling and evaluating a long lateral section as single oil producer targeting a carbonate reservoir. While no tar presence was expected, a combination of density, neutron porosity and nuclear magnetic resonance (NMR) logs while drilling resulted in identifying a deficit NMR porosity when compared to density porosity. Deployment of a formation pressure testing while drilling (FPWD) tool enabled measurement of the formation mobility and validate the presence of a tar. Using the same combination of measurements in the subsequent wells for delineating the tar enabled accurate planning of injection wells on the periphery of the field. Approximately 3 days were saved compared to the first well where the drill string had to be POOH to run-in with FPWD service. Hence, having FPWD tool in the same string helped in confirming the formation mobility in real time to call for critical decision making like changing the well trajectory or calling an early TD. Across all the wells drilled in this field, the formation pressure, mobility and porosity measurements provided valuable input for optimum ICD placement and design. Successful identification of unexpected tar resulted in substantial rig time savings, accurate planning of asset utilization and added confidence in design and placement of lower completions by utilizing LWD data. Benefits of integrated data and services combination became clear for applications involving advanced reservoir characterization and enhanced well placement in complex carbonate reservoirs. From the offset wells, a tar was seen in deeper formations but the integration of LWD NMR and mobility data from this well confirmed the presence of a tar within the zone of interest. The study established a cost-effective workflow for mitigating uncertainties related to tar encountered while drilling extreme ERD laterals in an offshore environment where any lost time results in significant increase in expenditures during the development phase. A systematic approach to tackle these uncertainties along with acquisition of critical data for the design & placement of completion results in optimum production from the reserves.

Root Cause Analysis of Wells with Integrity Failures

A Root Cause Analysis (RCA) is a methodological process of problem solving. In brief, this is an approach of "post mortem" analysis of the consequences with aim to understand what is one single (or multiple) lack of Management System that led to failure. Subsequently we can develop detailed remedial plan and actions to address failed Management System and most importantly prevent reoccurrence. Approach has been widely used in science and engineering. Probably, will be difficult to identify inventors of this analysis however first appearance in engineering discipline credited to Sakichi Toyoda, founder of Toyota Industries. He improved RCA by implementing technique called the "5 whys". Despite obvious benefit and versatility of root cause analysis methodology there are several challenges that might jeopardize result:–Absence of critical data / information due to various reason (time gap, no recoverable samples etc)–Too many variables that not allow to pinpoint main line / chain of investigation–Multiple failures with different root causes–"Depth of investigation" how many "Whys" are efficient to reveal main root cause.

Research Note: School Reopenings During the COVID-19 Pandemic and Implications for Gender and Racial Equity

In the fall of 2020, school districts across the country reopened under a variety of instructional modes. Some districts returned to in-person instruction and some operated remotely. Others reopened under hybrid models, wherein students alternated times, days, or weeks of in-person instruction. To capture this variation, we developed the Elementary School Operating Status (ESOS) database. ESOS provides data on elementary school districts' primary operating status in the first grading period of the 2020–2021 school year, covering 24 million students in more than 9,000 school districts in all states. In this research note, we introduce these data and offer two analytical examples. We show that school districts with greater representation of Black and Hispanic students were less likely to offer in-person instruction than were districts with greater representation of White students. These racial disparities remained after accounting for geographic locale and COVID-19 prevalence. We also show that the number of in-person elementary school instruction days was associated with mothers' labor force participation relative to fathers and to women without children—that is, the fewer days of instruction, the less likely that mothers were employed. ESOS is a critical data source for evaluating the mid- and long-term implications for students who experienced reduced in-person learning and for mothers who exited employment in the absence of in-person instruction and care.