Anytime Ranking on Document-Ordered Indexes

Inverted indexes continue to be a mainstay of text search engines, allowing efficient querying of large document collections. While there are a number of possible organizations, document-ordered indexes are the most common, since they are amenable to various query types, support index updates, and allow for efficient dynamic pruning operations. One disadvantage with document-ordered indexes is that high-scoring documents can be distributed across the document identifier space, meaning that index traversal algorithms that terminate early might put search effectiveness at risk. The alternative is impact-ordered indexes, which primarily support top- disjunctions but also allow for anytime query processing, where the search can be terminated at any time, with search quality improving as processing latency increases. Anytime query processing can be used to effectively reduce high-percentile tail latency that is essential for operational scenarios in which a service level agreement (SLA) imposes response time requirements. In this work, we show how document-ordered indexes can be organized such that they can be queried in an anytime fashion, enabling strict latency control with effective early termination. Our experiments show that processing document-ordered topical segments selected by a simple score estimator outperforms existing anytime algorithms, and allows query runtimes to be accurately limited to comply with SLA requirements.

Automated food intake tracking requires depth-refined semantic segmentation to rectify visual-volume discordance in long-term care homes

Malnutrition is a multidomain problem affecting 54% of older adults in long-term care (LTC). Monitoring nutritional intake in LTC is laborious and subjective, limiting clinical inference capabilities. Recent advances in automatic image-based food estimation have not yet been evaluated in LTC settings. Here, we describe a fully automatic imaging system for quantifying food intake. We propose a novel deep convolutional encoder-decoder food network with depth-refinement (EDFN-D) using an RGB-D camera for quantifying a plate’s remaining food volume relative to reference portions in whole and modified texture foods. We trained and validated the network on the pre-labelled UNIMIB2016 food dataset and tested on our two novel LTC-inspired plate datasets (689 plate images, 36 unique foods). EDFN-D performed comparably to depth-refined graph cut on IOU (0.879 vs. 0.887), with intake errors well below typical 50% (mean percent intake error: $$-4.2$$ - 4.2 %). We identify how standard segmentation metrics are insufficient due to visual-volume discordance, and include volume disparity analysis to facilitate system trust. This system provides improved transparency, approximates human assessors with enhanced objectivity, accuracy, and precision while avoiding hefty semi-automatic method time requirements. This may help address short-comings currently limiting utility of automated early malnutrition detection in resource-constrained LTC and hospital settings.

An Innovative Method to Extract Data in a Real-time Data Warehousing Environment

ETL (Extract, Transform, and Load) is an essential process required to perform data extraction in knowledge discovery in databases and in data warehousing environments. The ETL process aims to gather data that is available from operational sources, process and store them into an integrated data repository. Also, the ETL process can be performed in a real-time data warehousing environment and store data into a data warehouse. This paper presents a new and innovative method named Data Extraction Magnet (DEM) to perform the extraction phase of ETL process in a real-time data warehousing environment based on non-intrusive, tag and parallelism concepts. DEM has been validated on a dairy farming domain using synthetic data. The results showed a great performance gain in comparison to the traditional trigger technique and the attendance of real-time requirements.

Detecting Teeth Defects on Automotive Gears Using Deep Learning

Gears are a vital component in many complex mechanical systems. In automotive systems, and in particular vehicle transmissions, we rely on them to function properly on different types of challenging environments and conditions. However, when a gear is manufactured with a defect, the gear’s integrity can become compromised and lead to catastrophic failure. The current inspection process used by an automotive gear manufacturer in Guelph, Ontario, requires human operators to visually inspect all gear produced. Yet, due to the quantity of gears manufactured, the diverse array of defects that can arise, the time requirements for inspection, and the reliance on the operator’s inspection ability, the system suffers from poor scalability, and defects can be missed during inspection. In this work, we propose a machine vision system for automating the inspection process for gears with damaged teeth defects. The implemented inspection system uses a faster R-CNN network to identify the defects, and combines domain knowledge to reduce the manual inspection of non-defective gears by 66%.

DISTRIBUTIONS OF AIR TRAFFIC CONTROL STUDENTS’ ATTITUDES TOWARDS WORKLOAD

Mental workload is a well-known concept with a long development history. It can be used to examine students’ attitudes at the end of the educational process and compare them in groups or separately. However, building a continuous workload profile across the range of task complexity increase is still an urgent issue. All four groups of methods used to define mental workload have such flaws for the workload profile construction process as significant time requirements, single value processing and post-processing of the received results. Only one of them can be used without modifications to construct the operator’s attitude chart (profile) regarding the workload range and it doesn’t operate with more reliable absolute values. To resolve this problem, a special workload assessment grid was developed, considering the advantages of a subjective group of methods and seven core characteristics. The reasoning for grid axes choice, threshold values, and question formulation were provided. Statistics were calculated for the full sample, different grades, and educational institutions. Comparison of the received responses with referential values, cross-comparison between institutions and different grades were performed. The results contribute to such important aspects of workload, as redlines, workload profiling, and operator’s comparison.

When Digital Twin Meets Network Softwarization in the Industrial IoT: Real-Time Requirements Case Study

The Industrial Internet of Things (IIoT) is known to be a complex system because of its severe constraints as it controls critical applications. It is difficult to manage such networks and keep control of all the variables impacting their operation during their whole lifecycle. Meanwhile, Digital Twinning technology has been increasingly used to optimize the performances of industrial systems and has been ranked as one of the top ten most promising technological trends in the next decade. Many Digital Twins of industrial systems exist nowadays but only few are destined to networks. In this paper, we propose a holistic digital twinning architecture for the IIoT where the network is integrated along with the other industrial components of the system. To do so, the concept of Network Digital Twin is introduced. The main motivation is to permit a closed-loop network management across the whole network lifecycle, from the design to the service phase. Our architecture leverages the Software Defined Networking (SDN) paradigm as an expression of network softwarization. Mainly, the SDN controller allows for setting up the connection between each Digital Twin of the industrial system and its physical counterpart. We validate the feasibility of the proposed architecture in the process of choosing the most suitable communication mechanism that satisfies the real-time requirements of a Flexible Production System.

Low-Temperature Metallization and Laser Trimming Process for Microwave Dielectric Ceramic Filters

This paper describes a low-temperature metallization and laser trimming process for microwave dielectric ceramic filters. The ceramic was metalized by electroless copper plating at a temperature lower than those of conventional low-temperature co-fired ceramic (LTCC) and direct bond copper (DBC) methods. Compared with filters made via traditional silver paste sintering, the metal in the holes of the microwave dielectric filters is uniform, smooth, and does not cause clogging nor become detached. Further, the batches of fabricated filters do not require individual inspection, reducing energy, labor, cost, and time requirements. A microwave dielectric filter was then manufactured from the prepared ceramic using a laser trimming machine with a line width and position error within ±50 μm; this demonstrates a more accurately controlled line width than that offered by screen printing. After using HFSS software simulations for preliminary experiments, the microwave dielectric filter was tuned to a target Wi-Fi band of 5.15–5.33 GHz; the return loss was <−10 dB, and the insertion loss was >−3 dB. To implement the real-world process, the laser parameters were optimized. Laser trimming has a higher success rate than traditional manual trimming, and the microwave dielectric filter manufactured here verified the feasibility of this process.

Factorisation Path Based Refactorisation for High-Performance LU Decomposition in Real-Time Power System Simulation

The integration of renewable energy sources into modern power systems requires simulations with smaller step sizes, larger network models and the incorporation of complex nonlinear component models. These features make it more difficult to meet computation time requirements in real-time simulations and have motivated the development of high-performance LU decomposition methods. Since nonlinear component models cause numerical variations in the system matrix between simulation steps, this paper places a particular focus on the recomputation of LU decomposition, i.e., on the refactorisation step. The main contribution is the adoption of a factorisation path algorithm for partial refactorisation, which takes into account that only a subset of matrix entries change their values. The approach is integrated into the modern LU decomposition method NICSLU and benchmarked against the methods SuperLU and KLU. A performance analysis was carried out considering benchmark as well as real power systems. The results show the significant speedup of refactorisation computation times in use cases involving system matrices of different sizes, a variety of sparsity patterns and different ratios of numerically varying matrix entries. Consequently, the presented high-performance LU decomposition method can assist in meeting computation time requirements in real-time simulations of modern power systems.

Time and Side-Effect Burden Associated with Red Blood Cell Transfusions in Patients Diagnosed with Transfusion-Dependent β-Thalassemia

Introduction: β-thalassemia is a genetic blood disorder marked by the inability to produce hemoglobin. Diagnosed primarily in childhood, transfusion-dependent β-thalassemia (TDT) is the most severe form of this disorder and results in chronic anemia, requiring frequent red blood cell (RBC) transfusions to maintain hemoglobin levels. While previous studies have shown that health-related quality of life (HRQoL) is adversely affected in patients with TDT, the issues associated with the transfusion process itself in these patients (eg, side effects and time requirements) have not always been examined. These issues are the focus of this study. Methods: Following exploratory qualitative interviews with a small sample of patients identified by the Cooley's Anemia Foundation (CAF), a 30-minute web-based survey was developed to assess HRQoL variables in TDT. Subsequently, CAF identified patients in the US who were ≥ 18 years of age, had a self-reported physician diagnosis of β-thalassemia, and had received ≥ 1 RBC transfusion in the past 6 months, and invited them to participate in the study. Following informed consent, patients completed a number of validated and ad hoc questions about their treatment, side effects, psychological burden, time requirements for a transfusion, and caregiver information. Data were collected in May 2021. Descriptive results are presented as frequencies, means, and standard deviations (SD). Results: A total of 100 patients were identified and recruited by CAF and completed the survey. Of these, 35 (35%) were male, with a mean age of 36.0 (SD = 10.4) years. Patients had been diagnosed with TDT for a mean of 34.4 years (SD = 10.3) and had received a mean of 9.6 (SD = 4.3) RBC transfusions in the previous 6 months. In addition to receiving RBC transfusions (100%), other treatments ever received for TDT included iron chelation therapy (94%), surgery (47%), and stem cell/bone marrow transplantation (5%). Regarding RBC transfusions, 66% of patients reported requiring them more than once per month, 31% each month, and 3% every other month. Common patient-reported side effects of RBC transfusions were iron overload (81%), rashes/hives (74%), fatigue (61%), and pain/bruising at the injection site (52%). Additionally, 86% of patients reported feeling fatigued leading up to RBC transfusions, 75% reported that treatments for TDT had become routine, and 64% accepted the routine. Time required for an RBC transfusion was estimated at a mean of 7.4 hours (SD = 11.3) including transfusion preparation time. Travel and wait times were estimated as requiring an additional 7.1 hours. Other time requirements included frequent outpatient visits and frequent laboratory tests, averaging 3.9 visits and 6.1 visits in a 6-month period, respectively. Overall, 13% of patients reported having a caregiver. Caregiver activities included travel to and from appointments (77%), help with household duties (77%), and assistance in treatment decisions (62%). Conclusions: The results of this survey suggest there is a substantial time and side-effect burden associated with transfusion and treatment among patients diagnosed with TDT. This includes considerable time devoted to preparing for and receiving transfusions, transfusion-related side effects, frequent laboratory and outpatient appointments, and help from a caregiver. Medical treatments that would decrease the dependency on transfusions in TDT could decrease this burden and improve overall quality of life. Disclosures Gupta: Kantar Health: Current Employment; Bristol Myers Squibb: Consultancy, Research Funding. Costantino: Kantar Health: Current Employment; Bristol Myers Squibb: Consultancy, Research Funding. Perkowski: Kantar Health: Current Employment; Bristol Myers Squibb: Consultancy, Research Funding. Inyart: Kantar Health: Current Employment; Bristol Myers Squibb: Consultancy, Research Funding. Ashka: Kantar Health: Current Employment; Bristol Myers Squibb: Consultancy, Research Funding. Clapp: Kantar Health: Current Employment; Bristol Myers Squibb: Consultancy, Research Funding. Price: Bristol Myers Squibb: Current Employment, Current equity holder in publicly-traded company. Knoth: Bristol Myers Squibb: Current Employment.

Zero-Shot Pipeline Detection for Sub-Bottom Profiler Data Based on Imaging Principles

With the increasing number of underwater pipeline investigation activities, the research on automatic pipeline detection is of great significance. At this stage, object detection algorithms based on Deep Learning (DL) are widely used due to their abilities to deal with various complex scenarios. However, DL algorithms require massive representative samples, which are difficult to obtain for pipeline detection with sub-bottom profiler (SBP) data. In this paper, a zero-shot pipeline detection method is proposed. First, an efficient sample synthesis method based on SBP imaging principles is proposed to generate samples. Then, the generated samples are used to train the YOLOv5s network and a pipeline detection strategy is developed to meet the real-time requirements. Finally, the trained model is tested with the measured data. In the experiment, the trained model achieved a [email protected] of 0.962, and the mean deviation of the predicted pipeline position is 0.23 pixels with a standard deviation of 1.94 pixels in the horizontal direction and 0.34 pixels with a standard deviation of 2.69 pixels in the vertical direction. In addition, the object detection speed also met the real-time requirements. The above results show that the proposed method has the potential to completely replace the manual interpretation and has very high application value.